Membrane Hsp70—A Novel Target for the Isolation of Circulating Tumor Cells After Epithelial-to-Mesenchymal Transition

The presence of circulating tumor cells (CTCs) in the peripheral blood is a pre-requisite for progression, invasion, and metastatic spread of cancer. Consequently, the enumeration and molecular characterization of CTCs from the peripheral blood of patients with solid tumors before, during and after treatment serves as a valuable tool for categorizing disease, evaluating prognosis and for predicting and monitoring therapeutic responsiveness. Many of the techniques for isolating CTCs are based on the expression of epithelial cell surface adhesion molecule (EpCAM, CD326) on tumor cells. However, the transition of adherent epithelial cells to migratory mesenchymal cells (epithelial-to-mesenchymal transition, EMT)—an essential element of the metastatic process—is frequently associated with a loss of expression of epithelial cell markers, including EpCAM. A highly relevant proportion of mesenchymal CTCs cannot therefore be isolated using techniques that are based on the “capture” of cells expressing EpCAM. Herein, we provide evidence that a monoclonal antibody (mAb) directed against a membrane-bound form of Hsp70 (mHsp70)—cmHsp70.1—can be used for the isolation of viable CTCs from peripheral blood of tumor patients of different entities in a more quantitative manner. In contrast to EpCAM, the expression of mHsp70 remains stably upregulated on migratory, mesenchymal CTCs, metastases and cells that have been triggered to undergo EMT. Therefore, we propose that approaches for isolating CTCs based on the capture of cells that express mHsp70 using the cmHsp70.1 mAb are superior to those based on EpCAM expression.


INTRODUCTION
The malignancy of solid tumors is determined by the heterogeneity of the individual tumor (1) and its capacity to disseminate from the primary tumor to secondary sites via the blood stream. This process involves fundamental activities including invasion of the extracellular matrix and surrounding stroma, migration, evasion from the vasculature, proliferation, and the avoidance of anoikis (a form of programmed cell death that occurs in anchorage-dependent cells when they detach from the surrounding extracellular matrix) (2). Due to multiple selection steps, primary tumors and corresponding metastases often differ in molecular markers and pathways (3). Interrogating the biology of CTCs and also circulating tumor cell aggregates, which are generally considered as precursors of metastasis (4), is likely to provide invaluable insights into the biology of the disease and the prevalence and heterogeneity of future metastasis and thereby help to predict outcome and identify new tumor biomarkers that could be targeted by individualized antitumor therapies (5). Isolation, enumeration and the molecular characterization of CTCs could therefore bridge the "missing link" between cancer biology and individualized/precision therapies. As a liquid biopsy, CTCs do not only provide an opportunity to identify therapeutic targets and resistance mechanisms and monitor therapeutic responsiveness, but can also provide prognostic information about the risk of developing metastatic relapse and tumor progression which underpin 90% of cancer-related deaths (6)(7)(8).
A major limitation to the application of CTC-based analysis is the rarity of this cell type in the peripheral blood. Typically, 1 ml of peripheral blood of patients with metastatic cancer contains <10 CTCs (9-11). As a consequence, different strategies including filtration- (12,13), microfluidic chip- (14)(15)(16)(17), PCR- (18)(19)(20), and flow cytometry-based techniques (21) have been developed to enrich, separate and differentiate CTCs in the peripheral blood from the vastly more prominent hematopoietic cell population (1 CTC in 10 6 -10 8 hematopoietic cells). Although a variety of biomarkers such as CD44, CD133, CD47, cMET, EGFR, and immune checkpoint inhibitors are heterogeneously expressed on CTCs derived from different tumor entities, the most common techniques for the ex vivo separation of CTCs from peripheral blood are based on the capturing of cells using antibodies directed against cell surface expressed EpCAM (CD326) (22)(23)(24)(25)(26). The CellSearch R system (27)-the FDAapproved "gold standard"-combines a magnetic separation technique based on EpCAM antibody-coated particles with subsequent cytokeratin (CK) staining and a microscopic analysis of the isolated cells (22). Another limitation of most ex vivo CTC isolation techniques is the relatively small blood sample volume (7.5 ml) which is used and the low numbers of CTCs that can be derived therefrom. To overcome these disadvantages of ex vivo CTC isolation, GILUPI GmbH (Potsdam, Germany) has developed an EpCAM antibody-coated CellCollector R system Abbreviations: CK, cytokeratin; CTC(s), circulating tumor cell(s); EMT, epithelial-to-mesenchymal transition; EpCAM, epithelial cell adhesion molecule; Hsp70, heat shock protein 70; mAb, monoclonal antibody; mHsp70, membranebound form of Hsp70 which involves the direct insertion of a stainless steel wire, functionalized with gold and a hydrogel coating that incorporates anti-EpCAM antibodies, into the blood stream via a standard venous cannula in the cubital veins for 30 min. During this period, CTCs can be captured from the entire peripheral blood compartment (several liters of blood) of a cancer patient. Subsequently, the captured viable cells can be stained whilst attached to the wire and analyzed by fluorescence microscopy (28) or expanded for further analysis. The number of CTCs captured by the CellCollector R system before and after therapy has been shown to be associated with prognosis and therapeutic responsiveness (11).
All the techniques described above rely on the cell surface expression of EpCAM and the lack of the leukocyte marker CD45 by CTCs. However, many studies have shown that the transition of the adherent epithelial cells to the migratory mesenchymal state which enables the motility and invasiveness of CTCs and their dissemination to distant sites is associated with a loss in the expression of classical epithelial cell markers, including EpCAM (29). Yu et al. demonstrated that benign and non-invasive tumor cells exclusively express epithelial antigens, whereas a subpopulation of invasive breast cancer cells express both epithelial and mesenchymal markers (30). Epithelial-tomesenchymal transition (EMT) correlates with an increased migratory and metastatic potential of CTCs, invasiveness, poor overall survival and drug resistance (29,30). It is therefore apparent that systems for isolating CTCs that rely only on the expression of epithelial markers by target cells are limited in their ability to detect CTCs arising after EMT.
The search for universal tumor markers has revealed that the major stress-inducible heat shock protein 70 (Hsp70) is frequently expressed on the plasma membrane of primary tumor cells and distant metastases (31). This membrane Hsp70 (mHsp70) positivity has been identified on a large variety of different primary tumor types such as breast, lung, head and neck, colorectal, pancreas, brain and hematological malignancies, but not on corresponding normal cells and tissues (32,33). A comparison of the cell surface density of Hsp70 has also revealed higher intensities of mHsp70 on metastases compared to corresponding primary tumors in mouse and human models (33)(34)(35)(36). This finding provides a first indication that the expression of mHsp70 might not be downregulated by EMT and that it could therefore serve as a useful target for the isolation of CTCs in the circulation that have undergone EMT. Given that our group has developed a unique mouse monoclonal antibody (mAb) termed cmHsp70.1 which specifically detects the membrane-bound form of Hsp70 on viable tumor cells (37), herein we determine the capacity of the cmHsp70.1 mAb to form the basis of improved bead-and wire-based CTC isolation techniques that exploit mHsp70 expression as a universal tumorspecific biomarker.

Ethics, Patient Characteristics
Signed informed consent was obtained from all patients with squamous cell carcinoma of the head and neck (SCCHN) and non-small cell lung carcinoma (NSCLC) before EDTA blood (1-2 × 7.5 ml) was taken and the protocol was approved by the institutional ethical review board of the Klinikum rechts der Isar at the Technische Universität München (TUM). Eight patients with SCCHN in tumor stages pT1 to pT3, one patient with cancer of unknown primary (CUP) tumor of the head and neck, and three patients with advanced adenocarcinoma of the lung (NSCLC) were included into the study. All cells were maintained under standard conditions (37 • C, 95% v/v humidity, 5% v/v CO 2 ) in appropriate cell culture medium. Cell line authentication was performed by DNA profiling using highly polymorphic short tandem repeats (DSMZ-Deutsche Sammlung von Mikroorganismen und Zellkulturen, Leipzig, Germany (TUM) and the ATCC Cell Line Authentication Service (NTU). Cells were passaged every 3-4 days and used in the exponential growth phase. Cell viability was confirmed prior to use using trypan blue dye exclusion (>95%). Cell cultures were routinely tested to ensure the absence of mycoplasma.

Influence of TGFβ-Induced Epithelial-to-Mesenchymal Transition (EMT) on mHsp70 and EpCAM (CD326) Expression
EMT was induced in DU145 prostate, UP154, and UD5 squamous cell carcinoma cells of the head and neck and A549 lung cancer cells by treatment with TGFβ (38). For this, after initial seeding, cells were allowed to adhere for 24 h, after which the medium was exchanged with medium containing 10 ng/ml TGFβ (PeproTech) every second day. Control cells were sham treated in standard cell culture medium. After 10 days of cell culture with regular media changes, cells were harvested and the expression of mHsp70 and EpCAM on viable cells was determined by 2-color flow cytometry, as detailed below.
Influence of L-Lactic-Acid-Induced Epithelial-to-Mesenchymal Transition (EMT) on mHsp70 and EpCAM (CD326) Expression In addition to TGFβ, EMT was induced in A549 lung cancer cells, LS174T melanoma cells and UP154 and UD5 squamous cell carcinoma cells of the head and neck by a treatment with L-lactic-acid (Lac-Ac). For this, after initial seeding, cells were incubated with medium containing L-lactic-acid (Lac-Ac, 10 mM; pH 6.8). Up to a concentration of 20 mM, L-lactic-acid did not induce significant cell death in tumor cells. After 2 days, when cells reached a confluency of 70-80% cells were harvested and analyzed for the expression of EpCAM (CD326) and mHsp70 by flow cytometry. Control cells were sham treated with standard culture medium.
At Nottingham Trent University, the co-expression of mHsp70 and EpCAM on A549 and DU145 cells that had been induced to undergo EMT by treatment with TGFβ was determined by incubating cells (
In order to simulate EMT in vitro, prostate (DU145), head and neck (UP154, UD5) and lung (A549) cancer cells and malignant melanoma (LS174T) cells were incubated with TGFβ for 10 days and with L-lactic-acid (Lac-Ac, 10 mM, pH 6.8) for 2 days. Although TGFβ treatment drastically reduced the expression of EpCAM by DU145 cells (64 ± 5% to 18 ± 1%) (Figure 2A) and A549 (62 ± 1% to 2 ± 1%) cells (Figure 2B), mHsp70 expression was retained at nearly 100% after the induction of EMT in both cell lines (Figures 2A,B). Similar results were observed after a treatment of UP154 and UD5 head and neck cancer cells with TGFβ for 10 days (data not shown) and L-lactic-acid for 2 days (Figure 2A).
A comparative analysis of the mHsp70 and EpCAM expression on non-adherent and adherent A549 (Figure 2B) cells revealed a higher expression of EpCAM on the adherent,  epithelial-like cell population (A549, EpCAM: 64%) compared to the non-adherent, mesenchymal-like (A549, EpCAM: 46%) tumor subpopulation that dropped significantly upon treatment with TGFβ and L-lactic-acid ( Figure 2B). Concomitant with the mesenchymal transition induced by TGFβ and L-lactic-acid, the number of viable, non-adherent cells in the A549 cultures increased 15-and 11-fold, respectively. Similar results were observed for the LS174T cells. Following treatment with TGFβ and L-lactic-acid, the proportion of EpCAM positive cells decreased from 100 to 68% and 24%, respectively, whereas the proportion of cells expressing mHsp70 slightly increased from 76 to 82% and 89%, respectively ( Figure 2B).
Representative photomicrographs illustrating morphological changes induced by treating A549 cells with TGFβ for 10 days followed by a recovery period in the absence of TGFβ for 0 (d10), 4 (d10 + 4 days recovery), and 7 (d10 + 7 days recovery) days are provided in Figure 2C. Following treatment with TGFβ, a large proportion of cells detaches from the culture flasks and the remaining adherent cells show a spindle-like phenotype. Within the recovery period of 4 and 7 days, tumor cells became adherent and showed an epithelial-like phenotype. Similar results were observed after treatment with L-lactic-acid (data not shown).

Recovery of "Spiked" Tumor Cells by Magnetic Bead-Based Separation System Depends on the Proportion of mHsp70 Positive Cells
To assess and compare the capacity of cmHsp70.1 mAb coated magnetic beads to separate CTCs from buffer and peripheral blood and its relationship with mHsp70 expression, magnetic beads covalently coupled to cmHsp70.1 mAb were incubated with buffer and blood from healthy volunteers that had been "spiked" with identical numbers of SK-BR-3 breast cancer cells (∼93% positive for mHsp70 expression) or T47D breast cancer cells (∼23% positive for mHsp70) expression for 1 h at 37 • C.
As shown in Figure 3, following magnetic separation, the recovery of SK-BR-3 cells was significantly greater than the recovery of T47D cells (mHsp70: 80 ± 11% vs. 38 ± 13% and 83 ± 5% vs. 34 ± 7%, respectively) for both experiments (n = 3) in buffer ( Figure 3A) and blood ( Figure 3B). Capture therefore correlated with the percentage of mHsp70 positive cells in the respective tumor cell populations.

Vybrant TM CFDA-Stained and Unstained Tumor Cells Can Be Recovered With cmHsp70.1 mAb-Functionalized CellCollector ® System
The primary limitation of most ex vivo CTC isolation techniques is the relatively small volume of blood which is used (7.5 ml) and the low numbers of CTCs that can be derived therefrom. The GILUPI CellCollector R directly captures tumor cells from the patient's blood stream using an EpCAM antibody-coated CellCollector R system inserted into the cubital veins over a 30 min period. To evaluate the capacity of adapting this approach to capture cells expressing mHsp70, the cmHsp70.1 mAb was covalently linked to the surface of the detector tip of the CellCollector R system and this was then incubated with a suspension of Vybrant TM CFDA pre-stained or unstained SK-BR-3 or T47D cells in buffer. The capture of cells was then determined using a fluorescence microscope. As shown in Figure 4A, the cmHsp70.1 mAb-functionalized CellCollector R system captured more Vybrant TM CFDA pre-stained SK-BR-3 cells than T47D tumor cells. Similar results were obtained using unstained tumor cells that were subsequently fixed and stained with a cocktail of cytokeratin antibodies directly on the cmHsp70.1 mAb-functionalized CellCollector R system (data not shown). Representative fluorescence micrographs of SK-BR-3 and T47D tumor cells on the cmHsp70.1 mAb-functionalized wires after staining with DAPI and CK-FITC are provided in Figure 4B.

Tumor Cells Collected With cmHsp70.1 mAb-Functionalized CellCollector ® System From the Blood Maintain Their Phenotype and Can Be Propagated in Cell Culture
Circulating tumor cells (CTCs) are very rare and further analysis of these might require the in vitro expansion of isolated cells. The capacity to transfer cells that have been captured using the cmHsp70.1 mAb-based CellCollector R system into cell culture was therefore evaluated. For this, the cmHsp70.1 mAb-functionalized CellCollector R system was incubated with peripheral blood which had been spiked with unstained SK-BR-3 breast cancer cells under sterile conditions. After washing, the functionalized part of the detector was placed in a cell culture flask with appropriate culture medium. After 24 h at 37 • C, SK-BR-3 cells became adherent to the culture flask and started to grow. Flow cytometric analysis revealed that the phenotype of SK-BR-3 cells that were collected using the cmHsp70.1 mAb-based CellCollector R wire system was comparable to that of SK-BR-3 cells in cell culture (mHsp70: 94% vs. 91%; EpCAM: 100% vs. 100%, respectively) ( Figure 5).

cmHsp70.1 mAb-Functionalized Magnetic Beads Can Isolate CTCs From EDTA Blood of Patients With Squamous Cell Carcinoma of the Head and Neck (SCCHN) and Non-small Cell Lung Carcinoma (NSCLC)
Given the rarity of CTCs, it is also important to confirm that cmHsp70.1 mAb-functionalized magnetic beads can isolate CTCs from the peripheral blood of patients with cancer. We obtained EDTA blood from 8 patients with SCCHN, 1 patient with CUP tumor of the head and neck, and 3 patients with advanced NSCLC at diagnosis. The age of the SCCHN patients ranged between 51 and 72 years and that of NSCLC patients between 71 and 76. The clinical parameters of all patients are summarized in Table 1. Using magnetic bead separation with cmHsp70.1 mAb-functionalized magnetic beads we succeeded to isolate CTCs from 7 of 8 patients with SCCHN and from 3 of 3 patients with advanced NSCLC. The number of CTCs isolated from 7.5 ml EDTA blood of patients which are shown in Table 1 ranged between 0 and 92 cells. Representative fluorescence microscopic views of singular and clustered CTCs derived from equal amounts of EDTA blood (7.5 ml) of patients with SCCHN and NSCLC isolated with EpCAM mAb and cmHsp70.1 mAb-functionalized beads that were counter-stained with DAPI, CK-FITC, EpCAM-PE are shown in Figures 6A,B. cmHsp70.1 mAb-functionalized bead-separated, cytokeratin-positive CTCs were either EpCAMpositive or EpCAM-negative ( Figure 6A). In all tested cases, higher CTC numbers could be isolated from 7.5 ml EDTA blood of patients with SCCHN and NSCLC after separation with cmHsp70.1 mAb-functionalized magnetic beads compared to that isolated with EpCAM mAb-functionalized beads ( Table 1). Growth of isolated CTC clones was observed 14 days after   limiting dilution cloning of primary CTCs in 10 of 11 patients with SCCHN and NSCLC.
Future studies are planned to analyze the molecular characteristics of mHsp70 positive CTCs in a larger cohort of patients with different tumor stages, before therapy and in the follow-up period.

DISCUSSION
Tumor heterogeneity presents a significant barrier to the effective treatment of cancers in general and aggressive, therapy-resistant metastatic cancers particularly. Circulating tumor cells (CTCs) present in the peripheral blood offer an invaluable liquid biopsybased approach for interrogating and categorizing disease, evaluating prognosis and monitoring therapeutic responsiveness in patients with different cancer entities. Crucially, as the number of CTCs correlates with the stage of metastasis and is inversely related to therapeutic outcome, determining the presence, and interrogating the biology of CTCs provide the insights into the development and mechanism(s) of metastatic spread (39) and therapeutic resistance. Both are essential if progress in reducing cancer mortality is to be made, given that 90% of cancer-related deaths are due to therapy-resistant metastatic disease rather than the primary tumor (40). The ability to phenotypically and genetically profile CTCs from individual patients would provide an unprecedented insight into the biology of the disease on an individual basis and would inform and underpin the more effective delivery of precision-based, individualized medicine.
Currently, existing antibody-based approaches for isolating CTCs use magnetic beads or particles for the capture and isolation of CTCs. Cell-Search R , the current FDA-approved "gold standard" (22) and the GILUPI CellCollector R (28) rely on the cell surface expression of EpCAM (CD326) on the CTCs that are to be captured (41). However, the expression of EpCAM is often downregulated on CTCs (2), and is also known to be downregulated following EMT, a process which is integral to the transition of adherent tumor cells to a migratory status which enables them to depart from the primary tumor, enter the circulation and seed distal sites (2). As cells that have undergone EMT are those that are mostly involved in the establishment and progression of metastatic disease, it is essential that strategies for detecting, isolating and subsequently characterizing these CTCs are based on approaches that can best capture these cells. It is therefore essential that tumor markers that are expressed on CTCs before and after EMT are used as targets for capturing CTCs.
The search for universal tumor markers has revealed that the major stress-inducible Hsp70 is frequently expressed on the plasma membrane of a wide variety of tumor entities, but not non-transformed cells and tissues (31)(32)(33). It has also been demonstrated that metastases often express higher levels of mHsp70 compared to the primary tumor (33)(34)(35)(36). It is therefore likely that CTCs will preferentially express mHsp70 over EpCAM.
We have previously reported on the development and validation of a mouse monoclonal antibody (cmHsp70.1) which is able to bind to this membrane form of Hsp70 (37). The unique binding characteristics of this antibody make it an ideal candidate for the development of new bead-or wire-based approaches for capturing CTCs that will not be detected using EpCAM-based approaches, due to the downregulation of EpCAM expression after EMT. Herein, we profiled the expression of mHsp70 and EpCAM by cancer cell lines derived from a range of different tumor entities and the influence of TGFβ-and L-lactic-acidinduced EMT on expression and examined the capacity of bead-and wire-based approaches for capturing cancer cells differentially expressing mHsp70.
Flow cytometric profiling revealed heterogeneous mHsp70 and EpCAM expression patterns in cancer cell lines, and it is expected that the level of heterogeneity within and between different tumor entities will be even more marked in the clinical setting. Furthermore, we have demonstrated that although TGFβ-and L-lactic-acid-induced EMT results in a loss of EpCAM expression, mHsp70 expression is retained. The recovery of cancer cell lines expressing high and low levels of mHsp70 from buffer and blood obtained from health donors into which cancer cell lines had been "spiked" using cmHsp70.1 mAb-functionalized magnetic beads and a modified version of the GILUPI CellCollector R incorporating cmHsp70.1 mAb-mediated capture revealed that, as expected, the yield of recovered cells closely correlated with the level of mHsp70 expression of the respective cell line.
Since CTCs are typically very rare−1 ml of peripheral blood of patients with metastatic cancer contains <10 CTCs (9-11, 42)-further analysis and experimentation requires an ability to expand isolated cells in vitro (43). We could show that it is possible for CTCs isolated from patients with SCCHN and NSCLC to be transferred into cell culture after cmHsp70.1 mAb-mediated capture. Patient-derived primary CTCs isolated by cmHsp70.1 mAb-functionalized magnetic beads were either EpCAM-negative or EpCAMpositive, indicating that a proportion of EpCAM-negative CTCs cannot be captured using EpCAM-based isolation methods.
In summary, herein we have shown that cmHsp70.1 mAb-functionalized bead-and wire-based approaches provide a promising strategy for the detection and isolation of CTCs from the blood of patients with cancer. This is especially pertinent and important for the capture of CTCs that have undergone EMT, as EMT is associated with a downregulation of EpCAM, the target antigen on which most antibody-based approaches for capturing CTCs currently depend. This new approach could therefore improve the capacity to interrogate the presence and biology of CTCs and form the basis of a new liquid biopsy-based strategy for categorizing disease, evaluating prognosis and for predicting and monitoring therapeutic responsiveness across a range of aggressive cancers.

AUTHOR CONTRIBUTIONS
SB, SS, CW, WS, DL, GAF, and SW performed experiments. GM and AGP designed study and wrote the manuscript. AP and KK provided samples and clinical data from patients with SCCHN and NSCLC. GP collected, processed, and analyzed samples.