Information

At what cancer stage do tumors release circulating tumor cells into the blood?


And would a very accurate sensitive system for detecting circulating tumor cells (which detects 1 cell per 50 billion) be useful as a screening tool ?


Tumour cells can also metastasise using the lymphatic system. They first need to detach (produce enzymes that break down the extracellular matrix or any adherent molecules etc). Then they need to survive in the blood which is really hard with all the mechanical stress of being pushed around, the completely different nutrients, the different levels of blood gases, the immune cells in the blood and numerous filter mechanisms, then to truly metastasise they need to settle down somewhere and grow in their new environment. That's a lot of things right? So a sensitive assay like one you describe would need to have that tumour cell in the sample (unlikely, and only one cell is required to make a new tumour), is likely to be unable to tell that the cell is a tumour cell, wouldn't be able to differentiate between tumour cells that could or couldn't survive in the blood or differentiate those that could then go on to grow more in a tissue or not. And of course you're missing the entire lymphatic system. Following all of that, why does it matter? See a cancer, cut it out if you can and give chemotherapy to get rid of any possible cells outside of what you cut. If your assay came back negative or positive, treatment wouldn't change.


Screening programs are ruled out following sharp guidelines, like the so-called Wilson' criteria and subsequent ones provided by WHO. Specific documents regarding cancer screening have been published. A good primer on what a screening program in medicine is, can be found on wikipedia.

Conventionally, a tumor is considered non capable of metastasis (neither through blood, nor through lymphatic vessels), unless it doesn't invade the basement membrane, that is a thin layer of tissue that separates epithelias from the underlying tissues. In invasive neoplasiae, the surgeon have to cut it until the resection margins are free from tumour cells. This is achieved through real time microscopical examination by a pathologist during the procedure. If an ideal resection can't be performed, often medical therapy (e.g. chemotherapy) is suggested to the patient as neoadjuvant therapy. A neoadjuvant therapy is given in order to transform non optimal resectable tumors in good surgically treatable ones.

Commercial kits like this 1 , 2 , 3 , 4 are available on the market for this kind of test. Although there are this kit to find this cells, this method is not suitable for screening.


How blood flow helps cancer to spread

Metastasis is the spread of cancer to other parts of the body and the main reason why the disease is so serious. Now, brand new research reveals that blood flow is a key factor in this process.

Share on Pinterest What role does blood play in the spread of cancer?

In a paper that has now been published in the journal Developmental Cell, the scientists — who are from the National Institute of Health and Medical Research in France — describe their tests on zebrafish and humans.

The experiments confirmed that blood flow influences the locations at which migrating cancer cells “arrest” inside blood vessels.

They also detail how these cancer cells exit through the blood vessel walls and set up secondary tumor sites.

“A long-standing idea in the field,” explains senior study author Dr. Jacky G. Goetz, head of the laboratory at the University of Strasbourg in France — where the study was conducted — “is that arrest is triggered when circulating tumor cells end up in capillaries with a very small diameter simply because of size constraints.”

However, as Dr. Goetz explains, their findings show that “physical constraint” is not the only driver of metastasis, because “blood flow has a strong impact on allowing the tumor cells to establish adhesion with the vessel wall.”

Metastasis is the process through which tumor cells depart and migrate from their primary sites and travel through the lymph system or bloodstream to establish secondary, or metastatic, tumors in distant parts of the body.

Metastasis is a leading cause of cancer death and of “primary importance in the prognosis of cancer patients.”

It is a complex process and proceeds as a sequence of steps, each of which must be completed in order for the secondary tumor to flourish. The series of steps, known as the “metastatic cascade,” proceeds as follows:

  1. invading nearby healthy tissue
  2. crossing the walls of neighboring blood vessels and lymph nodes
  3. traveling through the bloodstream or lymph system to distant parts of the body
  4. arresting in remote, small blood vessels, or capillaries, invading their walls, and crossing over into the surrounding healthy tissue
  5. seeding a viable, tiny tumor in the healthy tissue
  6. generating a dedicated blood supply by growing new blood vessels to feed the new tumor

The new study concerns the fourth step, in which circulating tumor cells arrest in a capillary and cross through their endothelium, or the barrier of cells that line the vessel walls, into the surrounding tissue.

In their study paper, the authors explain that “very little is known about how [circulating tumor cells] arrest and adhere to the endothelium of small capillaries and leave the bloodstream by crossing the vascular wall.”

An area that is particularly unclear, they add, is the “role played by mechanical cues encountered in the blood” during this step.

For their study, the scientists developed “an original experimental approach” in which they tagged and followed circulating tumor cells as they traveled through blood vessels in zebrafish embryos. The model also allowed them to vary and measure blood flow in the vessels.

The results showed that the locations in the blood vessels at which the circulating tumor cells stop traveling is closely linked to flow rates.

The authors note that the “threshold velocity value for efficient adhesion […] ranges from 400 to 600 [micrometers per second].”

The team also found that blood flow is essential for “extravasation,” the process through which the tumor cells exit the blood vessels.

This was evident in timelapse imaging that showed endothelial cells “curling” around the arrested tumor cells in the zebrafish embryos’ blood vessels.

“ Blood flow at this step is essential. Without flow, endothelial remodeling does not occur. You need a certain amount of flow to keep the endothelium active so that it can remodel around the tumor cell.”

Dr. Jacky G. Goetz

The researchers came to the same results when they observed the progress of brain metastases in mice.

For this experiment, they used an imaging technique called intravital correlative microscopy, which combines living cell models with electron microscopy so that the dynamics can be observed in a live animal.

Finally, the team confirmed the findings by observing secondary tumors in the brains of 100 human patients, whose primary tumors were in various parts of the body.

As with the zebrafish model, they used an imaging technique to map the locations of the secondary tumors.

When they merged the brain metastases map with a blood flow map of a healthy control patient, the researchers found that it matched what they found in the zebrafish model, confirming that secondary tumors prefer to grow in areas where the blood flow is within a certain range.

The authors conclude that their findings reveal that blood flow controls not only the location, but also the onset of “metastatic outgrowth.”

They now want to explore ways to block endothelium remodeling around the circulating tumor cell as a way to disrupt its exit into surrounding tissue. Such an achievement could prevent metastasis from completing the steps necessary for the successful growth of a secondary tumor.


Circulating Tumor Cells

David G. Hicks MD , Susan C. Lester MD, PhD , in Diagnostic Pathology: Breast (Second Edition) , 2016

Prognostic Value

Newly diagnosed breast cancer patients may demonstrate CTCs in peripheral blood ○

Detection of CTCs associated with worse prognosis, independent predictor of progression-free and overall survival –

Prognostic association of CTC is independent of nodal status or whether patient receives adjuvant chemotherapy

Studies have defined > 5 CTCs per 7.5 mL of blood as cutoff for CTC(+) associated with unfavorable prognosis

Metastatic breast cancer patients with CTCs have worse prognosis than those without elevated CTCs ○

In 1 study, patients who were negative for CTCs showed median overall survival of 28.3 months –

Patients with elevated CTCs had median overall survival of only 12.8 months

CTC enumeration correlates with disease progression up to 7-9 weeks prior to imaging evidence of progression

Changes in CTC during therapy can serve as early parameter to evaluate response

Prognostic value of CTCs may depend on breast cancer subtype

Detection Methodologies for Circulating Tumor Cells

FeaturesImmunologic DetectionRT-PCR Detection
Detection methodologyImmunohistochemistry or immunofluorescencePCR for RNA transcripts
Detection targetsEpithelial antigens, cytokeratin (may use cocktails), HER2Cytokeratin, MUC1, mammaglobin, CEA, HER2 transcripts
SensitivityLess sensitiveMore sensitive
SpecificityMore specific (ability to view captured cells)Less specific (no ability to visualize captured cells)
Detection limits∼ 1 tumor cell per 10 6 nontumor cells∼ 1 tumor cell per 10 7 nontumor cells
CostMore expensive (requires image analysis equipment)Less expensive
Quantitative resultsYes: Yields results in terms of number of tumor cellsYes: Yields results in terms of measured number of gene transcripts, may vary widely between cells and between tumors

CTC isolation technologies

CTCs are known as an important marker for auxiliary diagnosis, prognosis evaluation, treatment decision, etc. To further extend CTCs’ clinical application, it is necessary to develop specific and effective techniques to capture rare CTCs from peripheral blood. Here we generally classify all CTC isolation techniques into biological and physical methods according to their enrichment principles (Fig. 1).

A mind map summarizing CTC isolation technologies. GEDI: geometrically enhanced differential immunocapture GO: graphene oxide VerIFAST: vertical immiscible filtration assisted by surface tension ISET: isolation by size of epithelial tumor cells FMSA: flexible micro spring array DFF: Dean Flow Fractionation p-MOFF: parallel multi-orifice flow fractionation MOFF-DEP: multi-orifice flow fractionation and dielectrophoresis

Biological isolation methods

Biological isolation methods are characterized by using specific surface markers, such as EpCAM. CellSearch is the gold standard for CTCs, capturing cells with specific EpCAM. The MagSweeper system introduces EpCAM-modified immunomagnetic beads, which are suitable for isolating circulating endothelial progenitor cells (CEpCs) with low to medium EpCAM expression. The three generations of the CTC-chip were developed to show increasingly higher isolation efficiency on CTCs, providing CTC samples with higher quality. The NanoVelcro chip is characterized by using specific antibody-modified nanomaterial substrate. One disadvantage of above methods is that they cannot effectively isolate CTCs with non-specific surface antigen expression. To overcome this defect, scientists are exploring new methods, even combining biological and physical isolation together, and achievements like CTC-iChip have been made (Additional file 1: Table S1).

Physical isolation methods

Physical isolation methods are based on CTC physical properties such as size (microfilter), membrane charge (dielectrophoresis), and density (density gradient centrifugation), etc. The combination of physical properties with some specific platforms, such as microfluidics, also shows great potential in capturing CTCs. Most of these methods do not require specific surface markers on CTCs. These techniques are generally simple in principle but must depend advanced materials or assistive engineering technologies for better clinical application (Additional file 1: Table S1).


Results:

Between March 10, 2014 and April 10, 2018, 586 patients underwent eligibility assessment (Figure 1). Among the 217 participants included, 7 patients withdrew their consent after randomization. We finally included 210 patients for the intention-to-treat analysis (107 in the sevoflurane group 103 in the propofol group).


Figure 1. Flow chart

The baseline characteristics are shown in Table 1. The baseline circulating tumor cell counts and positivity (using cut-off values ​​of at least 1 and at least 5 circulating tumor cells/7.5ml of blood) were similar in the two groups. The intraoperative and postoperative characteristics of each group are shown in Table 2.

See Figure 2 and Table 3 for the circulating tumor cell count over time. The type of anesthesia does not affect the circulating tumor cell count (median circulating tumor cell count/7.5ml blood [interquartile range]: propofol anesthesia group: 0 hour 1[0-4], 48 hour 1[0-2] , 72 hours 0[0-1] Sevoflurane anesthesia group: 0 hours 1[0-4], 48 hours 0[0-2], 72 hours 1[0-2] rate ratio, 1.27[95% CI, 0.95-1.71] P=0.103). However, the administration of sevoflurane anesthesia resulted in a significant increase in the maximum circulating tumor cell count after surgery (sevoflurane vs propofol: rate ratio, 1.36 [95%CI, 1.18-1.56] P<0.0001 that is, with propofol Compared with the use of sevoflurane, the maximum circulating tumor cell count increased by 1.36 times).

In the exploratory model, the effect of inhalation anesthesia on the maximum circulating tumor cell value after surgery is still significant (sevoflurane vs propofol: rate ratio, 1.26 [95% CI, 1.09-1.47] P=0.002 adjust tumor type , Size and dosage of opioids).


Table 1. Baseline characteristics


Table 2. Intraoperative and postoperative features


Figure 2. Circulating tumor cell count over time


Table 3. Circulating tumor cell counts during the perioperative period

Sixty patients (30 in the sevoflurane group and 30 in the propofol group) were randomly selected for in vitro exploratory analysis. The two groups of natural killer cells induced similar apoptosis rates (the average apoptosis rate in the sevoflurane group was 34.7% 35.7% in the propofol group). There is no evidence of linear regression that there is an association between the apoptosis rate and the maximum circulating tumor cell count (regression coefficient, -0.077 95% CI, -0.33-0.17 Figure 3).


Figure 3. Scatter plot of natural killer cell activity and maximum circulating tumor cell count after treatment


3. CTC as a dynamic prognostic factor in metastatic BC

3.1. Clinical validity in metastatic BC

In 2004, a first study reported a significant clinical validity of CTC count in metastatic BC with the CellSearch ® system (Cristofanilli etਊl., 2004). One hundred seventy‐seven patients starting a new line of treatment had their CTC count assessed at baseline and after few weeks of treatment. A threshold to distinguish patients with short versus long‐progression free survival was investigated, set up at 𢙕 CTC/7.5 ml of blood on the basis of a training cohort and confirmed in a validation cohort. This prognostic value was also observed during treatment (Hayes etਊl., 2006). By combining dichotomized CTC count (high or low) at two time‐points (baseline and after 1 cycle of treatment), 4 different progression𠄏ree survival profiles were obtained. The worse prognosis was seen in patients with high CTC count at both time points, as expected. Interestingly, patients with a high CTC count at baseline but a low CTC count after one cycle of therapy had a much better prognosis, almost similar to that of patients with low CTC count at baseline. Together with analytical validity data, these clinical data of CTC count as a dynamic prognostic biomarker prompted the FDA to clear the CellSearch ® technique as an 𠇊id in the monitoring of patients with metastatic breast cancer”. The initial claim was that CTC count monitoring during therapy would allow early detection of resistance to therapy and ultimately improve the management of metastatic BC patients.

Several reports were then published, most of them of limited size (level of evidence II–III) and with some discordant results. The IC 2006� study was the first observational study specifically designed and powered (267 patients) with CTC count as the primary endpoint (level of evidence 1 study) and confirmed most of the initial findings (Pierga etਊl., 2012a). Ten years after the seminal study, a pooled analysis of 1944 individual patient data finally established undisputable results on CTC in metastatic BC and demonstrated for the first time the superiority of CTC count over serum tumor markers (CEA, CA15.3). First of all, 𢙑 CTC can be detected in about 70% of stage IV BC patients and CTC count is associated with performance status, number of metastatic sites, elevated LDH, elevated serum tumor markers but not with tumor subtype. In addition, CTC count is a dynamic prognostic marker of progression𠄏ree survival and overall survival. Hazard ratio of survival between high and low CTC counts increases together with the threshold used to define high CTC count. Finally, in contrast to serum tumor markers, adding CTC count and its change during therapy to an optimized clinico‐pathological model significantly increases the prognostic value of the model (Bidard etਊl., 2014).

3.2. Clinical utility in metastatic BC

The first trial initiated to demonstrate the clinical utility of early CTC changes after one cycle of chemotherapy was conducted in the US by the SouthWest Oncology Group. In the SWOG S0500 trial metastatic BC patients whose CTC count stays above 5 CTC/7.5 ml of blood after the first cycle of the first line of chemotherapy were eventually randomized to an early switch to the second line of chemotherapy. Such switch was hypothesized to greatly improve the patients overall survival compared to the standard imaging�sed management (targeted hazard ratio: 0.6). The study results were published in 2014 and no survival difference was seen between the two arms (Smerage etਊl., 2014). To explain these negative results, it has been discussed by the study investigators that second line chemotherapy is unlikely to have a significant effect (even when introduced earlier on the basis of elevated CTC count) on BCs that have a primary resistance to first line chemotherapy. Other comments have been made on the trial's design and concepts (Alunni�roni etਊl., 2014 Bidard and Pierga, 2015). On the basis of these negative results, the 2015 American Society of Clinical Oncology clinical practice guidelines for CTC count considered reasonable for clinicians to not use CTC count in women with metastatic BC (Van Poznak etਊl., 2015). While this negative trial had a major impact on the use of CTC count by US clinicians, two other clinical utility trials based on CTC count are currently ongoing in France:

The 𠇌irCe01” trial ( <"type":"clinical-trial","attrs":<"text":"NCT01349842","term_id":"NCT01349842">> NCT01349842) is also based on the early changes of CTC count, but patients are enrolled before the start of third line of chemotherapy and followed with the CTC test throughout the successive lines of chemotherapy.

The “STIC CTC” trial ( <"type":"clinical-trial","attrs":<"text":"NCT01710605","term_id":"NCT01710605">> NCT01710605) investigates the clinical utility of the prognostic value of baseline CTC count. In this trial, the choice of the first line of treatment (hormone therapy or chemotherapy) for relapsing hormone‐positive BC is determined either by the clinician or by the baseline CTC count.

The study designs have been reported elsewhere (Bidard etਊl., 2013b), and the results are expected within the next 2 years.


Materials and Methods

PATIENT RECRUITMENT AND MATERIAL COLLECTION

A total of 115 men with increased serum PSA concentrations (>4 ng/mL) or abnormal findings on digital rectal examination suspicious for PCa were recruited for this study after written informed consent was obtained. The study was approved by the local ethical review board under number PV3779. TRUS-guided diagnostic core biopsies were performed between August 2012 and January 2016. Histologic diagnosis of 8 to 12 tissue cores was carried out according to the Gleason score by an experienced pathologist. Two CellSave Preservative tubes were each filled with 7.5 mL of peripheral blood, 1 before and 1 within 30 min after performing prostate biopsy. The tubes contained Na2EDTA for clotting prevention and a cell preservative to maintain the morphology and cell-surface antigen expression of epithelial cells for phenotyping. The samples were directly processed to detect CTCs using the CellSearch system. Researchers analyzing the samples were blinded to all patient-related data, including time point of blood sampling. Characteristics of the patient cohort are presented in Table 1. Patients with histological diagnosis of PCa were treated with radical prostatectomy, radiation therapy, or active surveillance.

Characteristics of the total cohort of participants. a

. PCa− (n = 40) . PCa+ (n = 75) . Total (n = 115) . P value .
Age at diagnosis, years (mean) 40–82 (61.4) 46–79 (66.1) 40–82 (64.5) 0.0086
Total PSA, ng/mL (median) 1.9–13.4 (7.1) 2.5–304.6 (7.7) 1.9–304.6 (7.7) 0.0586
Free PSA, ng/mL (median) 0.2–2.8 (1.1) 0.2–3.2 (0.9) 0.2–3.2 (0.9) 0.2438
Free PSA/total PSA ratio, % (median) 7.8–25.9 (14.9) 2.2–35.6 (10.8) 2.2–35.6 (12.9) <0.001
Gleason score
3 + 3 n = 24
3 + 4 n = 26
4 + 3 n = 5
≥4 + 4 n = 20
T stage
T1 n = 10
T2 n = 34
T3 n = 20
Treatment
Prostatectomy n = 46
Prostatectomy + other n = 7
Radiotherapy n = 13
Other/none n = 9
. PCa− (n = 40) . PCa+ (n = 75) . Total (n = 115) . P value .
Age at diagnosis, years (mean) 40–82 (61.4) 46–79 (66.1) 40–82 (64.5) 0.0086
Total PSA, ng/mL (median) 1.9–13.4 (7.1) 2.5–304.6 (7.7) 1.9–304.6 (7.7) 0.0586
Free PSA, ng/mL (median) 0.2–2.8 (1.1) 0.2–3.2 (0.9) 0.2–3.2 (0.9) 0.2438
Free PSA/total PSA ratio, % (median) 7.8–25.9 (14.9) 2.2–35.6 (10.8) 2.2–35.6 (12.9) <0.001
Gleason score
3 + 3 n = 24
3 + 4 n = 26
4 + 3 n = 5
≥4 + 4 n = 20
T stage
T1 n = 10
T2 n = 34
T3 n = 20
Treatment
Prostatectomy n = 46
Prostatectomy + other n = 7
Radiotherapy n = 13
Other/none n = 9

Significant test for mean: Welch 2-sample t-test significant test for median: Wilcoxon rank-sum test with continuity correction.

Characteristics of the total cohort of participants. a

. PCa− (n = 40) . PCa+ (n = 75) . Total (n = 115) . P value .
Age at diagnosis, years (mean) 40–82 (61.4) 46–79 (66.1) 40–82 (64.5) 0.0086
Total PSA, ng/mL (median) 1.9–13.4 (7.1) 2.5–304.6 (7.7) 1.9–304.6 (7.7) 0.0586
Free PSA, ng/mL (median) 0.2–2.8 (1.1) 0.2–3.2 (0.9) 0.2–3.2 (0.9) 0.2438
Free PSA/total PSA ratio, % (median) 7.8–25.9 (14.9) 2.2–35.6 (10.8) 2.2–35.6 (12.9) <0.001
Gleason score
3 + 3 n = 24
3 + 4 n = 26
4 + 3 n = 5
≥4 + 4 n = 20
T stage
T1 n = 10
T2 n = 34
T3 n = 20
Treatment
Prostatectomy n = 46
Prostatectomy + other n = 7
Radiotherapy n = 13
Other/none n = 9
. PCa− (n = 40) . PCa+ (n = 75) . Total (n = 115) . P value .
Age at diagnosis, years (mean) 40–82 (61.4) 46–79 (66.1) 40–82 (64.5) 0.0086
Total PSA, ng/mL (median) 1.9–13.4 (7.1) 2.5–304.6 (7.7) 1.9–304.6 (7.7) 0.0586
Free PSA, ng/mL (median) 0.2–2.8 (1.1) 0.2–3.2 (0.9) 0.2–3.2 (0.9) 0.2438
Free PSA/total PSA ratio, % (median) 7.8–25.9 (14.9) 2.2–35.6 (10.8) 2.2–35.6 (12.9) <0.001
Gleason score
3 + 3 n = 24
3 + 4 n = 26
4 + 3 n = 5
≥4 + 4 n = 20
T stage
T1 n = 10
T2 n = 34
T3 n = 20
Treatment
Prostatectomy n = 46
Prostatectomy + other n = 7
Radiotherapy n = 13
Other/none n = 9

Significant test for mean: Welch 2-sample t-test significant test for median: Wilcoxon rank-sum test with continuity correction.

CTC ANALYSIS

Blood analysis with the CellSearch system was performed within 96 h as previously described ( 18) using the CTC Kit according to the manufacturer's recommendations. Briefly, epithelial cells among the cells captured by anti–epithelial cell adhesion molecule (EpCAM) antibodies were detected by binding of antibodies C11 and A.53B/A2, directed against keratins 8, 18, and 19, and potentially also recognizing keratins 4 through 6, 10, and 13 ( 19, 20). An anti-CD45 antibody was used to define and exclude leukocytes. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). After enrichment and immunocytochemical staining, immunomagnetically labeled cells were kept in a strong magnetic field and scanned using the CellSpotter Analyzer (Menarini-Silicon Biosystems). Experienced researchers interpreted results of these analyses. A blood sample of 7.5 mL was considered positive with the identification of at least 1 keratin-positive, DAPI-positive, CD45-negative cell showing no morphological signs of apoptosis. This cutoff was adapted from our studies on early-stage breast cancer because so far no other prognostic cutoff has been determined by previous publications for early-stage PCa.

STATISTICAL ANALYSIS

Statistical analyses were performed with Matlab R2016a (The Mathworks), R version 3.3.3 (R Foundation for Statistical Computing), and In-Silico Online version 2.0 ( 21). Two-sample t-tests and Wilcoxon rank-sum tests were used to calculate the statistical difference between means and medians, respectively. To calculate the odds ratios (ORs) in CTC change, the CTC counts were normalized to the amount of blood volume and correlated to the following independent variables using a Poisson distributed generalized linear mixed-effect model, corrected for age, with patients nested in time point: (a) time point of blood collection (before/after biopsy), (b) presence of PCa cells in the biopsies (yes/no), and (c) interaction effects. Progression-free survival from date of biopsy was modeled using the Kaplan–Meier function, and univariable and multivariable hazard ratios (HRs) were calculated using Cox proportional hazards analysis. Patients undergoing active treatment (radical prostatectomy or radiation therapy without concurrent androgen deprivation therapy) were evaluated with respect to the effect of CTC increase after therapy (increase of at least 1 CTC) on biochemical recurrence for progression-free survival. Biochemical recurrence after radical prostatectomy was defined as a PSA increase >0.2 ng/mL and PSA nadir + 2 after radiation therapy. Covariates consisted of age, PSA, number of positive cores, biopsy Gleason score, and treatment (radical prostatectomy vs radiation therapy). For calling statistical significance, the α level of 0.005 that was recently proposed by Benjamin et al. to increase credibility was applied ( 22).


Characterization and molecular profiling of CTCs

We have discussed various CTCs enrichment techniques which are being used for isolation of CTCs from metastatic cancer patients. However, none of them has achieved much of quantitative success. The results have shown a great amount of variation from 10% to 90% of isolated CTCs and hence, it is crucial to analyze the collected cells for their quantity as well as their exact phenotype. A numerical indication of collected CTCs may not be able to reveal the true picture of the type of cells isolated from cancer patients. Similarly, tumor cells can undergo a variety of changes and be present in heterogeneous subpopulations. Hence, a mere number of CTCs can lead to faulty conclusions. Therefore, there is a need for true characterization of these isolated CTCs cells to come to logical conclusions. Molecular profiling of these isolated cells will crystallize the picture, as it reveals the true nature of the isolated CTCs cells.

A fundamental process in EMT, down-regulates E-cadherin, which can be attained by many transcriptional factors. [39] Most of the molecular markers that have been isolated for characterizing CTCs are EMT indicators. During EMT process, a metastatic cell goes through a lot of modifications at cellular and molecular levels and many genes undergo transcriptional alterations. [39] Some of these genes play a role in initiating the effect of EMT while others play a role in regulating and maintaining its transited state. The other factors like inflammatory cytokines and physical changes in the tumor microenvironment also play a role in EMT promotion. [39] TWIST1 and TWIST2 genes are most strongly expressed genes in EMT process which are responsible for inducing transformation alone or in co-operation with other factors such as TGFβ, Wnt, Notch, etc. [40] E-cadherin is one of the most important proteins for maintaining the epithelial nature of cells. Snail1 and Snail2 suppress the transcription of E-cadherin as well as Zeb1 and Zeb2 genes. This results into downregulation of E-cadherin, which leads to initiation of EMT process. [41,42] Other gate keeper’s genes of epithelial state, such as alpha and gamma catenins are also been down-regulated along with downregulation of E-cadherin in this process. [43,44]

Induction of certain mesenchymal characters during EMT process requires upregulation of two extracellular matrix proteins, that is, vimentin and fibronectin in these cells which escape the barriers of local tissue and proceeded for invasion. Similarly, other genes such as N-cadherin, CD44, intergrin β6 are also implicated for proper migration of these cells. [43-46] Even understanding the mutational changes, abnormal size, and characteristics of CTCs, scientists are still pondering over the fact that these cells are able to survive in an environment which is totally hostile for them. It is postulated that out of the several hundred CTCs shed by the tumor, only a few remain in the circulation. There are reports suggesting that CTCs bearing mutations, such as upregulation of CD47, help them in escaping attack by natural killer cells and macrophages. Similarly, downregulation of chaperone protein-calreticulin again helps them to dodge the immune system. [47,48] Schölch et al. [49] in their studies have referred this state as an “immune-evasive” to the period between EMT and MET in circulation. Thus, overall it seems that CTCs have very evolved mechanisms to maintain and express their invasive aggressive nature by surpassing the body’s natural immune system.


Introduction

Solid tumors can release a surprisingly high number of circulating tumor cells (CTC) everyday into the circulation (1). Although CTCs originating from primary tumors are considered transitional in the search for a new home, most of these cells are fated to die in circulation owing to mechanical and environmental trauma such as shear forces, oxidative stress, and attack by the immune system. In fact, only a small fraction of CTCs are capable of surviving, seeding distant organs, and eventually giving rise to overt metastatic disease. Most CTCs have a short half-life of less than 2.5 hours in circulation (2) and are apoptotic (3, 4). Therefore, only the CTCs with a survival advantage during their transit in the blood stream and a better potential for colonization in the distant sites can likely contribute to metastasis. To discover better prognostic and predictive markers of early metastatic recurrence and novel targets for its prevention and treatment, it is critical to identify and characterize the CTC population with highest metastatic potential. Recent preclinical and clinical studies suggest a link between CTC clusters and worse clinical outcomes (3, 5). CTC clusters are defined as groups of two or more aggregated CTCs found in the blood of patients with solid tumors (6). Despite the prognostic implications for CTC clusters, the molecular mechanisms responsible for their formation or dissemination and the pathways conferring their survival advantage and metastatic potential remain largely unknown. Here, we examine preclinical evidence on the sources of CTC clusters, potential mechanisms of CTC clusters formation (genesis), transit to distant sites (dissemination), their survival advantage, and increased metastatic potential. We also deliberate open questions (Table 1), unmet research needs, and future directions (in italics) to delineate the clinical significance of CTC clusters in the metastatic cascade. We finally discuss methods that are most common and clinically relevant or novel and promising for isolating CTC clusters and describe clinical evidence for their prognostic value.

Open questions in various areas of CTC cluster research

Sources of CTC clusters

Both primary and metastatic tumors may constitute the source of CTC clusters forming multidirectional transit routes (Fig. 1A). CTC clusters originating from a primary tumor could “self-seed” the original site or travel to distant sites of metastasis. CTC clusters arising from a (micro)metastatic site could return to the primary tumor site or the original (micro)metastatic site or could travel to another distant site of metastasis. To support this hypothesis, tumor self-seeding has been a well-accepted concept for CTCs in general (7). The self-seeding potential (7) and the oligoclonality of CTC clusters (5) were both demonstrated in the same mouse xenograft model of MDA-MB-231 LM2 cell line. Although CTC clusters were found to be a minority (2.6%) in the overall CTC population in this model, their calculated probability of metastatic formation was 50 times higher, as suggested by formation of dual-color metastasis from dual-color primary tumors (5). Although the study evaluated self-seeding concept only for the primary tumors, cross-seeding of primary and (micro)metastatic tumors can also be envisioned. This multisite exchange of CTC clusters may allow communication between each tumor site to collectively acquire capability of surviving the tremendous treatment pressures, and eventually promoting the tumor growth and progression.

Sources and potential mechanisms of CTC clusters origin. A, Multidirectional transit routes of CTC clusters. CTC clusters originating from a primary tumor could “self-seed” the original site or travel to distant sites of metastasis. CTC clusters arising from a (micro)metastatic site could return to the primary tumor site or the original (micro)metastatic site or could travel to another distant site of metastasis. This multisite exchange of CTC clusters may allow communication between each tumor site to collectively acquire capability to survive the tremendous treatment pressures, eventually promoting the tumor growth and progression. B, Origin of CTC clusters due to “cell jamming.” The “cell jamming” principle proposes that increasing confinement from the growing mass of tumor or higher density of extracellular matrix may promote grouping of the cells. In this context, higher mammographic density may facilitate CTC cluster formation. C, Orchestrated origin of CTC clusters through activation of tissue development and regeneration pathways. These pathways include (i) intact cell polarization (not graphically represented in the figure), (ii) acquired expression of cell surface proteases and various cell adhesion molecules, (iii) the presence of adherens junctions for which plakoglobin is an important mediator, and (iv) remodeling of tissue/tumor architecture to clear the track, which is facilitated by a keratin-14–positive leader cell. These mechanisms may in turn facilitate tumor cell cooperativity and their collective cell migration as CTC clusters.

Genesis of CTC clusters

The physiological events by which CTC clusters originate are still largely unknown. The concept that CTC clusters could form by intravascular grouping of single CTCs has been disproven by a recent study (5). However, that still does not address the question about the exact steps leading to CTC cluster formation. An emerging hypothesis is that CTC clusters are formed due to “cell jamming” (Fig. 1B). This principle proposes that increasing confinement from the growing mass of tumor or higher density of extracellular matrix (ECM) may control the mode of tumor cell dissemination. Higher ECM density was shown to shift the preference of mesenchymal tumor cells to collective invasion whereas lower ECM density was associated with single cell invasion in an in vitro model (8). This principle may also apply to mammographic breast density, which is an important prognostic factor for locoregional recurrence in early-stage breast cancer and of progression-free survival in metastatic breast cancer at initial diagnosis (9, 10). However, the influence of breast density on “cell jamming” and preference for CTC cluster formation is an open question, which can be investigated in animal models with collagen I defects, LOX-mediated collagen crosslinking, or CD36 repression (11–13).

A more strategic preparation for the tumor cells to form a cluster may involve activation of pathways involved in tissue development and regeneration. These molecular mechanisms may in turn facilitate tumor cell cooperativity and collective cell migration (14). A few decades ago preclinical studies focusing on wound healing and morphogenesis have shown that epithelial cell aggregates are capable of spreading movements in vitro and in vivo (14). These aggregates are able to migrate while maintaining cell–cell interactions (15). Some of the mechanisms proposed for this collective cell migration include (i) intact cell polarization, (ii) acquired expression of cell surface proteases and various cell adhesion molecules, (iii) presence of adherens junctions, and (iv) remodeling of tissue/tumor architecture to clear the track (Fig. 1C). Interestingly, some of these pathways seem to be also important for ECM-induced collective invasion (8). If present in CTC clusters, these processes may allow cell–cell coupling and multicellular organization and ultimately facilitate the formation of CTC clusters. More research is needed to address the exact cellular events leading to CTC cluster formation, with a possibility of multiple passive and active modalities supported by ECM, tumor microenvironment, and/or tumor cells themselves.

Similar to morphogenetic movements, collective movement occurs in many cancers in which cells are not completely de-differentiated (16). In this regard, collective cell migration may be led by a keratin 14-positive leader cell, which may create a path for the other tumor cells in its group through the surrounding tissue, in the blood stream, and potentially in the invaded site (Fig. 1C ref. 17). Interestingly, ECM-induced “cell jamming” is shown to also require track clearance by leader cells (8). Plakoglobin, which is involved in cell–cell junction and is highly expressed in CTC clusters compared to single CTCs, may also provide a preference for CTC clusters formation and integrity throughout their transit in the blood (5) (Fig. 1C). Keratin 14, plakoglobin, E-cadherin, and other epithelial cytoskeletal and adhesion proteins form the core of the machinery necessary for formation and dissemination of CTC clusters. Here, the unexplored research topics include the molecular processes by which the epithelial framework facilitates CTC cluster formation and its dependence on various mechanisms of cell invasiveness and migration.

Dissemination of CTC clusters

The access of CTC clusters into the blood stream could be made possible by the porous and leaky blood vessels formed within rapidly growing tumor masses, via hasty neoangiogenesis (Fig. 2A ref. 18). This supports the tumor self-seeding hypothesis to allow the entry of CTC clusters back to the original site. Conversely, a choreographed entry through invadopodia and macrophage-dependent transendothelial migration (19) may also be possible for CTC clusters to gain the access to circulation (Fig. 2A). The invadopodia are the protrusive and adhesive structures of cancer cells thought to arise in response to a range of signals primarily from tumor microenvironment. The proteolytic function of invadopodia through localized activity of matrix metalloproteases and their role in transendothelial migration of individual tumor cells are particularly important for metastasis. However, whether similar invadopodia-based and/or macrophage-dependent pathways are involved in intravasation of CTC clusters for dissemination remains to be explored. Future studies may also include capturing the live events of CTC clusters in transit by utilizing recent advances in the three-dimension real-time microscopy imaging in vivo (20).

Dissemination, survival advantage, and increased metastatic potential of CTC clusters. A, Dissemination of CTC clusters. CTC clusters may enter the circulation either by porous and leaky blood vessels formed by hasty neoangiogenesis required by the growing mass of tumor or by a choreographed entry through invadopodia and macrophage-dependent transendothelial migration, which has been demonstrated for single CTCs. When in circulation, a CTC cluster may act like a thrombus and get arrested in the small veins or capillaries. Here, they may find residence and give rise to overt metastasis. An (re)arrangement of the CTCs within a cluster in a linear fashion as a single-cell file may allow the grouped cells to pass through microvessels to reach more distant sites. B, Survival advantage and increased metastatic potential of CTC clusters. Persistent adhesion-dependent survival signals in CTC clusters can provide survival stimuli and thus contribute to effective metastatic spreading. Although epithelial cell–cell interactions may be important, cellular plasticity of CTCs within clusters have also been seen as in form of expression of EMT markers and presence of hybrid cells with both epithelial and EMT characteristics. Disaggregation of CTC clusters by uPA or plakoglobin knockdown or any other method may give rise to single CTCs. Although single CTCs may experience many survival challenges such as shear forces, environmental or oxidative stresses, and immune assault leading to apoptosis, CTCs within clusters may be shielded from them. This does not rule out the possibility that a few single CTCs may still be able to colonize a distant site. Cellular heterogeneity [such as undifferentiated vs. differentiated and epithelial vs. EMT] for homotypic clusters made up of only tumor cells and interaction between tumor cells and other nontumor cells (such as stromal or immune cells) for heterotypic clusters may offer competitive advantage for colonization at distant sites. Furthermore, activation of tumor dormancy program could favor formation of micrometastatic niche and escape from immunosurveillance at the distant sites.

Once in circulation, CTC clusters have slower flow rate than single CTCs within the blood vessels (21), which suggests its embolus/thrombus-like behavior. In support of this, administration of a thrombolytic agent called urokinase-type plasminogen activator (uPA) is effective at breaking down CTC clusters into single cells as well as modestly reducing the numbers of metastatic lesions and improving survival (21). Contrariwise, uPA expression is associated with enhanced tumor migration and invasion and higher rates of tumor progression and metastasis (22, 23), suggesting a differential role of uPA in early versus metastatic stage. Disaggregation of CTC cluster has also been reported with genetic knockdown of plakoglobin, which leads to their compromised metastatic efficiency in the animal models (5). The observation of reduced metastasis and improved survival by disrupting CTC clusters directly, either by systemic uPA administration or by genetic knockdown of plakoglobin (Fig. 2B), in animal models raises interesting questions: Can disaggregation of CTC clusters in circulation provide an advantage in preventing metastatic disease? What are the “druggable” targets that can dissociate CTC clusters? Are there any FDA-approved drugs that can disaggregate CTC clusters? In this regard, CTC clusters are often found to be associated with platelets (24, 25). It is not clear if platelets play any role in maintaining their integrity during the transit or their dissemination ability. Whether currently available anti-platelet agents can dissociate CTC clusters may be an interesting question to evaluate in preclinical models.

Slowly moving CTC clusters if arrested in small veins or capillaries may find residence and give rise to overt metastases from within the vessel (Fig. 2A). Because lung is the first organ encountered by these CTC clusters released from various organs, this potentially could be one of the mechanisms by which they are responsible for lung metastases. Indeed, a preclinical study has demonstrated that lung metastasis originates from the intravascular proliferation of endothelium-attached tumor cells rather than from CTCs that were able to extravasate and invade lung parenchyma (26). This phenomenon might also explain the formation of brain metastases despite the presence of an intact blood brain barrier. However, a deliberate movement of a group of tumor cells through microvessels is also possible due to CTC clusters organized in a linear arrangement of single-cell files (27, 28), which may also explain the trans-pulmonary passage of CTC clusters into other organs and rise of metastases in other distant site (Fig. 2A). What factors influence the structure of CTC clusters and possibly the site of colonization for CTC cluster is another important area of research.

Survival advantage for CTC clusters

CTC death may be in part due to the loss of adhesion-dependent survival signals leading to anoikis (29), which might explain apoptotic CTCs of epithelial phenotype (Fig. 2B refs. 30, 31). This supports the hypothesis that persistent epithelial cell–cell interactions in the form of clusters can provide survival stimuli and thus contribute to effective metastatic spreading (16). However, CTC clusters also express more mesenchymal versus epithelial markers compared to single CTCs (31). How would that provide additional survival advantage from anoikis during the transit is not known. Although epithelial-to-mesenchymal transition (EMT) in CTC clusters is counterintuitive because it is expected to result in high propensity of single cells, the cellular plasticity and cooperativity within a cluster may confer resistance to various stresses within the circulation (32). In support of this, the hybrid epithelial/mesenchymal state of CTC clusters has recently been described (Fig. 2B ref. 25) however, its direct association to increased survival advantage is less clear. The molecular mechanism(s) allowing the cells to have the plasticity is a highly important area of investigation as they may provide additional pharmacological targets for the prevention of metastasis.

Other conceivable mechanisms for the survival advantage of clusters include the cooperation between cells within CTC clusters shielding from shear forces, environmental or oxidative stresses, and immune assault (Fig. 2B). Indeed, tumor fragments are found to survive and grow better after they are transplanted (33) or injected (34, 35) into a new host. In this context, heterotypic clusters containing more durable stromal or immune cells aggregated with CTCs may provide additional advantage (36, 37). It can also be hypothesized that paracrine interactions between cells of various origin in heterotypic clusters may play a pivotal role in seeding of tumor clusters and in the evasion of immunosurveillance at the distant site, further providing a survival advantage (Fig. 2B). The mechanisms that regulate these interactions and their influence on survival advantage and colonization potential of CTC clusters also remains unknown.

A predominantly glycolysis-driven cell metabolism of a cancer cell allows the cells to survive in hypoxic conditions while being maintained in the tumor microenvironment. While in circulation, the oxygen deprivation may be even more severely restricted so that only the toughest cells survive. Significant work has been done to understand how cancer cell metabolism affects the tumor cell growth as well as its migratory or invasive capability (38, 39). The role of EMT in rewiring of the cancer cell metabolic network and, vice versa, the importance of metabolic reprogramming on EMT are also starting to be deciphered (40). On one hand, EMT controls the expression of genes involved in metabolic pathways such as glycolysis, lipid metabolism, mitochondrial metabolism, and glutaminolysis. However, deregulated expression of metabolic enzymes in these pathways promotes EMT. Although these pathways are shown to be relevant in migrating single CTC with mesenchymal characteristics, the differences in cancer cell metabolism between single CTCs and CTC clusters need to be investigated.

Enhanced metastatic potential of CTC clusters

The polyclonal tumor cell cooperativity and crosstalk within the network of migrating homotypic (made of only CTCs) or heterotypic clusters (made of CTCs and other stromal/immune cells) may facilitate stabilizing and initiating metastatic growth (27). For homotypic clusters, cellular heterogeneity (such as undifferentiated vs. differentiated and epithelial vs. EMT) may offer competitive advantage for colonization at distant sites (Fig. 2B). While expression of stem cell markers within some CTCs has been described, this remains an important area of future investigation for CTC clusters. Single cell analysis of stem cell marker expression and the localization of tumor-initiating cells within the cluster may be an important factor to examine. For heterotypic clusters, interaction between tumor cells and other nontumor cells (such as stromal or immune cells) may also be important for colonization and escaping immunosurveillance (Fig. 2B). Given the recent findings of the various tumor-suppressive/promoting roles played by different tumor-associated immune cells (41), identifying the nontumor cell types, especially immune cells, and their interactions within clusters will shed light on their biological functions and clinical significance in metastasis. With the latest advances in single-cell molecular profiling technologies, analysis of each cell within clusters may reveal mechanisms of cooperativity and define the roles of nontumor cells within CTC clusters.

The observation that CTC clusters are nonproliferative (42), which may allow their escape from the pressures of cytotoxic treatments during the transit and at the distant site, is of interest. This may reflect activation of tumor dormancy programs (43), which, in turn, could favor formation of micrometastatic niches and escape from immunosurveillance at the distant sites (Fig. 2B). The clinical relevance of this nonproliferative status of CTC clusters is an area of active research (3). Although CTCs have been reported in cancer survivors 7 to 22 years after their initial treatment (2), whether CTC clusters are present in these survivors and if their presence predicts imminent recurrence is unknown. Because the “nondormant” state of CTCs as assessed by the proliferation index predicts relapse in breast cancer patients (44), the assessment of this dynamic in CTC clusters may provide additional insight into the mechanisms of tumor progression. Recently, it was proposed that cancer of unknown primary may be explained by the presence of dormant CTCs forming a premetastatic niche and giving rise to a metastatic disease before the tumors at the primary site can be detected (45). It will also be important to determine the role of CTC clusters in this process. Furthermore, the mechanisms by which cluster-host cell interaction at the distant site can intervene and facilitate metastatic cascade also need further investigation.

Isolation and detection of CTC clusters

The major challenges in isolating CTC clusters are related to (i) their paucity as they are found in numbers as low as one cluster per over 10 7 leukocytes and 10 10 red blood cells, (ii) the potential for their dissociation during blood processing, and (iii) the variations in their physical, cellular, and molecular characteristics. A desired platform to isolate CTC clusters would be able to isolate live and intact CTC clusters of different size, shape, and composition independently of tumor-specific cell surface markers with reduced processing time, robust clinical feasibility, and demonstrated clinical validity in predicting prognosis in patients. While considerable numbers of platforms have been developed for CTC isolation in general, only some have shown the capacity to detect clusters of CTCs, with only a handful demonstrating the prognostic significance of CTC clusters in patients, as described in the section below. These include (i) immunomagnetic-based isolation methods, such as the CellSearch system, which is currently the only FDA-approved platform for the detection of CTCs as a prognostic marker in metastatic cancer patients (46) (ii) size-based filtration methods, such as the Isolation by SizE of Tumor cells (ISET refs. 47, 48) and (iii) microfluidic devices or chips, operating on various passive or active separation principles (49). These platforms are compared in Table 2 for their desired features for CTC clusters research, which highlights the need to develop a platform specifically for CTC cluster research.

Main features of common methods for the detection and study of CTC clusters

To date, microfluidic devices seem to be the most promising platform for isolating CTC clusters, as they offer several advantages such as (i) ability to process whole blood without the need for red blood cells (RBC) removal, which results in less potential of cluster dissociation from shear or centrifugation forces and faster processing time and (ii) collection of live CTC clusters. The main drawback of this platform has been the need for cell surface marker-based capture. To overcome this limitation, size-based isolation methods using spiral microfluidics have been optimized (50). These spiral systems have shown excellent recovery rates and very efficient depletion of white blood cells. Importantly, these devices can be produced at low cost and be easily operated, making them available for a widespread use. To customize selective capture of CTC clusters, a first generation of platform named Cluster-Chip was developed. Cluster-Chip used specialized bifurcating triangular micropillars acting as traps under low-shear stress to preserve CTC cluster integrity and demonstrated high efficiency capture of clusters in patients with metastatic breast or prostate cancer and melanoma (51). To overcome the physical limitation of Cluster-Chip platform that impact viability of CTC clusters, the same group has recently adapted a two-stage deterministic lateral displacement (DLD) approach in a continuous flow microfluidic device to sort clusters based on size and asymmetry from whole blood (52). Here, the first stage is designed to extract clusters based on size such that larger ones will be moved laterally, while smaller clusters and single cells will follow the streamlines through the device to arrive at the second stage, which captures smaller clusters based on asymmetry. Another microfluidic device purposely developed to isolate clusters of CTCs is a three-dimensional (3D) scaffold chip, which can efficiently capture clusters by combining specific antibody-dependent recognition and physical barricade effect of the 3D scaffold structure (53). Here, the scaffold is uniformly coated with thermosensitive gelatin hydrogel, which dissolves at 37°C, allowing gentle release of the captured cells, and thus assuring high viability for downstream applications, including cell culturing.

The future innovation in microfluidic approach to capture CTC clusters requires integration of multiple separation principles to cover the wide physical variations seen in this rare population and shortening of the processing time. An ideal platform will also have demonstrated clinical feasibility as well as validity by confirming the prognostic value of CTC clusters in cancer patients. Commercialization of this platform will also be critical to undertake multitude of future studies described above to uncover the biological and clinical roles of CTC clusters.

Clinical relevance of CTC clusters

Although it is still unclear which tumor or patient characteristics can predict the presence of CTC clusters, recent clinical studies have demonstrated the prognostic value of CTC clusters (Table 3). The presence and high numbers of CTC clusters at baseline have shown to be associated with shorter progression-free survival (PFS) and overall survival (OS) in patients with various types of solid tumors (3, 54–60). Moreover, platinum resistance has been observed in patients with primary or recurrent ovarian cancer with CTC clusters (61). Finally, cancer patients with persistence of CTC clusters after treatment initiation and with bigger CTC cluster size are shown to have shorter survival (PFS and/or OS refs. 30, 56, 62). These clinical findings not only suggest the prognostic value of CTC clusters but also emphasize their biological significance in tumor progression and treatment outcome.

Association of CTC clusters and clinical outcomes in cancer patients

Summary

Evidence so far supports a functional role of CTC clusters in surviving pressures of travelling through the bloodstream, such as anoikis, shear forces, and immune attack, as well as colonizing distant organs. The advantage may be offered by the composition and cooperativity among the CTCs within a cluster, compared to the single CTCs. However, much is still unknown about their genesis, transit, and settlement. The clinical data so far also indicate the prognostic value of CTC cluster analysis in predicting treatment resistance and survival outcomes in cancer patients. Nevertheless, the precise cellular and molecular mechanisms enhancing the metastatic ability of clusters remain unclear. A combination of microfluidic and computational simulation of cluster movement with real-time in vivo microscopy may help us understand the early events of CTC cluster formation and dissemination. In addition, comparing molecular profiling of single versus clustered CTCs may reveal the pathways responsible for their extended survival and drug resistance and define the roles of nontumor cells associated with CTCs. The biological studies to answer the open questions related to CTC clusters presented here will allow a deeper understanding of the role of CTC clusters in tumor progression, identify novel therapies, and eventually guide clinical studies for personalization of therapeutic decision-making.


Author information

Daphne W. Bell & Ulysses J. Balis

Present address: Present addresses: National Human Genome Research Institute/NIH Cancer Genetics Branch, Bethesda, Maryland 20892, USA (D.W.B.) Department of Pathology, University of Michigan Health System, Ann Arbor, Michigan 48109, USA (U.J.B.).,

Sunitha Nagrath and Lecia V. Sequist: These authors contributed equally to this paper.

Affiliations

Surgical Services and BioMEMS Resource Center, Massachusetts General Hospital, Harvard Medical School, and Shriners Hospital for Children, Boston, Massachusetts 02114, USA

Sunitha Nagrath, Daniel Irimia, Ulysses J. Balis, Ronald G. Tompkins & Mehmet Toner

Massachusetts General Hospital Cancer Center, Harvard Medical School, Boston, Massachusetts 02114, USA

Lecia V. Sequist, Shyamala Maheswaran, Daphne W. Bell, Lindsey Ulkus, Matthew R. Smith, Eunice L. Kwak, Subba Digumarthy, Alona Muzikansky, Paula Ryan & Daniel A. Haber