WO2011002649A1

WO2011002649A1 - Analysis of circulating tumor-related microparticles

Info

Publication number: WO2011002649A1
Application number: PCT/US2010/039628
Authority: WO
Inventors: Frank Coumans; Leon W.M.M. Terstappen
Original assignee: Veridex, Llc
Priority date: 2009-06-30
Filing date: 2010-06-23
Publication date: 2011-01-06

Abstract

The methods, apparatus and kits described in this invention are used to analyze circulating tumor related microparticles. Analysis can be performed with a number of platforms, including flow cytometry or the Cellsearch system. Enumeration of TMP' s have been shown to correlate with CTC count and, therefore, provide an alternative means to assess metastatic disease. Because of their frequency in blood compared to CTC's, TMP 's provide a more sensitive indicator of dissemination and a more specific means to monitor disease.

Description

ANALYSIS OF CIRCULATING TUMOR-RELATED MICROP ARTICLE S

Frank Coumans and Leon W.M.M. Terstappen

Cross Reference to Related Applications

The present application claims the benefit of priority of U.S. Provisional Application No.

61/221579, filed June 30, 2009, which is incorporated herein by reference.

Background of the Invention

Because of the costs and severity of the disease, cancer has become one of the major diseases challenging healthcare systems in countries throughout the world. In most cases, death from cancer is not caused by expansion of the primary tumor, but by dissemination of the disease or metastasis.

The current assessment of disease burden depends upon the use of expensive instruments such as X-rays, MRI, PET, and CT scans. Interpretation of the results is costly and needs to be performed by specialists where any changes in disease state assessed over a period of months.

With the ever improving therapies and increasing life expectancy of patients, the number of patients living with cancer is increasing. Therefore, an assessment whether or not a tumor is disseminated is extremely important in monitoring the progression of the disease and managing the health care of the patient.

Dissemination can be directly monitored through the cells shed from the tumor. These cancer cells become dangerous when the cells break away from the primary site and secondary sites elsewhere in the body. The cascade of events involved in the metastatic process leads to the formation of new (disseminated) tumors. The resulting metastasis interferes with the function of organs, uses up the body's resources and ultimately leads to death. Therefore, it is of the utmost importance for the management of the disease to know whether the tumor has disseminated and to monitor the dissemination.

In US Pat No. 6,190,870, CTCs are assessed as both live and dead cells, wherein "dead" includes the full range of damaged and fragmented cells as well as CTC-derived debris. The conventional density gradients used in these studies would lose damaged CTC that would be located in the red blood cell (RBC) layer. CTC debris that would be positively stained for cytokeratin may also have densities falling in the RBC, since most intracellular components have densities in the range of 1.15 to 1.3. In addition, some damage to intact CTC cells may have occurred during cytospin or subsequent processing.

US Pat No. 6,670,197 describes methods for binding fragments and debris to beads. The published application describes numerous possibilities for the density of fragments and debris of interest. Upon centrifugation, the beads will be located in a layer of RBC, because of the predetermined specific gravity (density) of the beads, coupled to fragments and debris. However, this system is dependent upon correctly binding fragments and debris to these beads. If any other sample component binds the beads, they may not appear in the desired location and subsequently will not be subject to analysis.

US patent application Ser No. 2005/0181463 describes the analysis of circulating tumor cells in conjunction with circulating clusters, fragments and debris. The application focuses on minimizing in vitro damage to captured cells to assess the clusters, fragments and debris occurring as a consequence of apoptosis, necrosis or damage from the immune response. The application does not consider a specific tumor cell related microparticles (TMP) as indicators of dissemination, based on size and labeled components.

Herein are described methods to diagnose, monitor, and screen for dissemination of disease based on tumor cell related microparticles found in a blood sample. Also provided is an apparatus and kits for assaying biological specimens using these methods.

Brief Description of the Invention

The methods, reagents, device and kits described in this invention are used to analyze tumor cell related microparticles (TMP). TMPs most likely occur from apoptosis of tumor cells and provide similar clinical significance as circulating tumor cells (CTCs). Because of their frequency in blood compared to CTCs, TMPs provide a more sensitive indicator of

dissemination. Enumeration of TMPs provide a statistically more accurate and sensitive test in early detection of dissemination, especially when the CTC frequency is on an order below 1 cell per 7.5 ml of blood. Analysis for TMP's can be performed with a number of platforms, including multiparameter flow cytometry, the Cellsearch® system (Veridex, LLC), or any appropriate fluorescent imaging system. Brief Description of the Figures

Figure 1 Circulating Tumor Cell identified by the CellSearch system. Figure 2 Objects identified as Circulating tumor Cell candidates by the Cellsearch system. The yellow box is 4μm x 4μm and serves to gauge the minimum size set as criteria for a CTC.

Objects 1 and 2 are clear CTCs. The majority of operators will classify objects 3-6 as CTCs and objects 7 and 8 are not classified as CTC. Green is Cytokeratin-PE, Purple is DAPI. Figure 3 Circulating Tumor Cells in 30 ml of blood before surgery of newly diagnosed colorectal cancer.

Figure 4 Cytokeratin PE image of TMP candidates indicated with arrows for a patient (Panel A) and control (Panel B).

Figure 5 Correlation between CTC and TMP 's in the blood of 176 metastatic prostate cancer patients before initiation of a new line of therapy. R2 = 0.66, slope = 17, intercept = 20.

Figure 6 Kaplan Meier estimates of overall survival of metastatic prostate cancer patients with <150 TMP's or >150 TMP's (Panel A) and < 5 CTCs or > 5 CTCs (Panel B).

Figure 7 Representation of TMP generation. Panel A shows an intact tumor. Panel B is the same tumor in an apopotic state. Panel C is the degradation into TMP. Figure 8 Representation of the construction of immunoactive beads A, B, C and the capture and release of EpCAM + cells and TMP's.

Figure 9 Flowthrough device for capture of EpCAM+ CTCs and TMP's. Panel A: device components; Panel B: flow of blood and capture of CTCs and TMP's; Panel C: release of CTCs and TMP's and trapping on filters; Panel D: release of filters for analysis.

Figure 10 Three images of the 8 CTC classes. Green represents the Cytokeratin-PE staining and the purple the DAPI staining. The black and white images show the corresponding CD45 staining. All classes except "CK+/CD45+" were CD45 negative. The top panel shows four classes that were assigned during review after the algorithm identified CK+ / DAPI+ events. The bottom panel shows classes assigned during review of CK+ events. For L-TCF the nucleus has to be larger than a 4 μm square, while for S-TCF the nucleus has to be smaller than a 4 μm square. For L-TMP the CK staining has to be larger than 4 μm square, while for S-TCF the CK staining has to be smaller than a 4 μm square. CK+/CD45+ are positive in both CK and CD45+ channels. Focus Artifact (FA) images have the appearance of one or more out of focus balls, with a dot (partially) enclosed by a circle.

Figure 11 Kaplan Meier plots of overall survival of CRPC patients after initiation of therapy for each of the CTC classes. The vertical markers represent censored patients. The number of patients in each group and the group boundaries are shown in the top right corner of each plot.

Detailed Description of the Invention

A more recent approach to monitoring metastasis incorporates a direct analysis of suspect cells in a blood sample (CellSearch System, Veridex LLC). In this system, enumeration of CTC requires a trained personnel for interpretation, but has limited sensitivity. The level of tumor dissemination is determined by enumerating circulating tumor cells (CTC) in the blood of patients with disseminated tumors. Immunomagnetic enrichment and fluorescent imaging of cells provides a detection limit of 1 CTC in 7.5 ml of blood (see US 6,365,362; US 6,645,731). The system has been shown to be clinically useful as a means for associating the presence of tumor cells in the blood with rapid disease progression and short survival in patients with metastatic carcinomas, (see UA 20070037173). For example in comparing the number of CTCs in a 7.5 ml sample of blood from a patient treated for hormone refractory prostate cancer with his survival, 88 patients (38%) with less than 5 CTC (Favorable), before or after therapy, lived significantly longer (> 26 months) than the 71 patients (31%) with 5 or more CTC (Unfavorable; < 6.8 months). These latter patients were on an ineffective therapy. Similarly when 26 patients (11%) converted from Favorable to Unfavorable CTC levels, the median survival was 9.3 months while the patients who converted from Unfavorable to Favorable CTC levels improved their survival to 21.3 months, implying effective therapy. Accordingly, the level of CTC provides the medical oncologist with the opportunity to check the effectiveness of the chosen treatment. While in the case of non-efficient treatment, the ability to switch to a different treatment can be made at an earlier time point.

The CellSearch system consists of reagent kits, an automated sample preparation device (CellTracks Autoprep®, Veridex LLC) and a semi-automated fluorescence microscope

(CellTracks Analyzer II®, Veridex LLC). The sample preparation device performs an immunomagnetic cell enrichment of CTC from 7.5 ml of blood. Specificity of the enrichment is obtained by coating ferrofluids with antibodies specific for the epithelial cell adhesion molecule (EpCAM) expressed only on cells of epithelial origin. The system stains the enriched sample with the nucleic acid dye DAPI, Phycoerythrin (PE) labeled antibodies identifying Cytokeratin specific for cells of epithelial origin and Allophycocyan (APC) labeled antibodies CD45 identifying cells of hematopoietic origin. The fluorescence microscope is equipped with a 1OX objective (NA 0.45), a mercury arc lamp, a CCD camera, a computer controlled filter wheel and an X, Y, Z stage. The set up takes a set of images of the DAPI, APC and PE fluorescence intensity that covers the complete surface of the analysis chamber. A computer program is used to identify locations that show both DAPI and PE fluorescence. It presents the images to an operator who decides whether a detected event is a CTC or not, using a standard set of criteria. An image of a CTC with a DAPI stained nucleus, cytoplamic Cytokeratin and no CD45 staining is illustrated in Figure 1. In the overlay, green represents the Cytokeratin and purple the DAPI staining. A definition of a CTC as Cytokeratin-PE+, CD45-, DAPI+ and larger than 4μm was set to determine the accuracy, precision, linearity, and reproducibility of the CellSearch system. Typical images of 8 CTC candidates presented by the computer to the operator for review are shown in Figure 2. All corresponding images did not show staining with CD45 indicating that they are not white blood cells. In the current Cellsearch system only objects that are larger in size than the 4μm x 4μm yellow boxes are classified by the operators as CTC. Objects 1 and 2 are typical DNA and Cytokeratin containing whole cells and are classified as CTC. More difficult to assign are objects 3-6 leading inevitably to errors in the outcome. Objects 7 and 8 are too small to be classified as CTCs and are the subject of the present invention.

It has been reported that in 35% of patients with clinically confirmed disseminated disease no CTC are detected. Even though these patients have relatively better prognosis, life expectancy is still short, suggesting that CTC escaped detection because of their low frequency. In newly diagnosed breast and colorectal cancer patients before surgery (Figure 3), the number of CTCs detected in a 30 ml blood sample increased with the stage of the disease. However, a substantial portion of the 30 ml samples did not contain CTCs, suggesting that even a 30 ml test volume is not large enough for some patients (de Groot MR, Croonen HM, Mastboom WJT, Vermes I, Tibbe AGJ, Tissing H, Terstappen LWMM; "Circulating tumor Cells (CTC) in newly diagnosed breast or colorectal cancer, Proc Ann Meet Am Soc Clin One, 25:18s, 2007).

Consequently in the CellSearch system, some patients with 0 CTC in 7.5 ml of blood may have a tumor already metastasizing.

Similarly, patients with hormone refractory prostate cancer show a decrease in survival at higher CTC numbers. However a large portion of patients in which 0 CTC have been identified in 7.5 ml of blood probably do have CTC in the total 5 liters of blood in the body as some of these patients have a low survival. So while all of these patients may have 0 CTCs in a blood sample, it is obvious that they have disseminated disease and have tumor cells present in their peripheral blood albeit at a very low frequency. The explanation resides in the statistical error inherent in the detection of small numbers. If, on average, one CTC is present in 15 ml of blood, then the probability to detect 0 CTC in a test volume of 7.5 ml is 60% with the probability to find 1 or 2 cells in the test volume is 30% and 7.5%, respectively. Accordingly, a test volume of 7.5 ml of blood may not contain CTC otherwise present in the blood.

Clearly, the current sensitivity of the CellSearch System is not sufficient, especially in these patients. Thus, there is a need for an even more sensitive system that is applicable across all stages of cancer.

The present invention provides for a simple blood test using small tumor-related micro particles (TMP's) found in a patient's blood sample. The test allows a clinician to rapidly assess treatment and decide whether to stop or change treatment. The test will directly lead to an improvement of the treatment and thus quality of life of the patient. Further, the cost for medical care will be reduced.

TMP's have been observed in blood of cancer patients. These TMP's are believed to originate from apoptosis or the death of tumor cells. Quantitative analysis of their presence in the blood provides an indicator of dissemination and clinical information in the diagnosis and treatment of disease. Because the frequency of these TMP's is an order of magnitude larger than the frequency of CTC in blood, enumeration of TMP's are statistically more accurate and sensitive in testing blood samples. The importance of TMP's is better appreciated in the early detection of dissemination of tumor cells when the CTC frequency is still below 1 cell per 7.5 ml ofblood.

Thus, there is a further need for increased specificity beyond the CellSearch system. The occurrence of false positives in CTC detection for patients is low in 7.5 ml ofblood. However from 344 blood samples, taken from patients with no cancer, 1 CTC was detected as a false positive in 22 of the samples (6%), and 3 CTC were detected as a false positive in 1 of the samples (0.3%). In all other non cancer patient samples 0 CTCs were detected. In addition, the influence of false positive CTC in patients with no cancer increases with larger blood volumes and, consequently, the need to increase specificity.

Finally, an operator independent classification of CTC would improve the accuracy of detection beyond that of the CellSearch system. The Cellsearch system relies upon the operator for the final classification of captured cells as CTCs. Thus, there is a potential for variation in the accuracy and further error in the system. A system that does not rely upon an operator for the classification of CTC is therefore highly desirable.

The enumeration of TMP provides a new system to detect disseminating tumors with a higher sensitivity, specificity and accuracy than CTC enumeration.

Tumor cell related microparticles (TMP) are small cytokeratin positive objects which were captured using EpCAM ferrofluid from 7.5 ml ofblood in patient samples. TMP were defined as all objects that stain with cytokeratin labeling, are NOT CD45 positive, and are less than 4μm (Figure 4). The arrows indicate the position of TMP candidates and the blow-up in Figure 4, Panel A reveals the resolution of the system and the actual size of the three TMP candidates. The same positions are then investigated for the corresponding DAPI and APC image. Any TMP will be negative for APC staining. Other parameters such as, but not limited to, DAPI staining, size, PE intensity and roundness are considered in the present invention as potentially relevant to clinical outcome and could be included. Together, these parameters can be incorporated in developing an algorithm to identify TMP 's.

When TMP are defined as Cytokeratin+, CD45- and < 4μm, application to a CellSearch dataset from a metastatic prostate clinical trial involving patients with no cancer versus metastatic patients with cancer and before initiation of therapy showed a correlation r of 0.66 with a slope of 17 and an intercept of 20 with the number of CTCs in the same patients. Figure 5 shows the frequency of TMP' s and their correlation with CTCs. Thirty seven (37) of 176 patients (21%) had 0 CTCs detected, 27 of the 37 had TMP 's above the intercept suggesting that a substantial portion of the patients with O CTCs can be further subdivided based on their TMP number. The higher the frequency could result in a more sensitivity and accurate test when correlated with clinical outcome.

Kaplan Meier plots showed the relationship between TMP 's and survival of the prostate cancer patients. As shown in Figure 6, patients were divided into a Favorable and Unfavorable group using the median of the TMP counts and the presence of 5 or more CTC in a volume of 7.5 ml of blood. The Kaplan Meier plots surprisingly show as good or even better separation of the Favorable and Unfavorable groups when counting TMP 's. While there is a higher background observed in the blood of patients with no cancer, it is expected that this background is in part due to the definition used to count the TMP 's. The present invention considers other definitions, especially relating to size, in the enumeration of TMP's that would correlate with survival.

TMP's appear to be very heterogeneous, where some types may have clinical

significance and others have no diagnostic meaning. Accordingly, the present invention considers microscopic, spectroscopic and cytometric means for characterizing TMPs and their clinical significance. TMP detection, together with patient history, provides a data base for correlating patient outcome and patient survival. The present invention further considers algorithms based on this type of data which could be incorporated into software, designed to count TMP's in patients and provide patient information for clinical use.

While not meaning to limit the scope of the present invention, one hypothesis for the origin of TMP's and their generation is shown in Figure 7. The generation of TMP's begins with an intact CTC having a cytoplasmic membrane, a cytoskeleton containing Cytokeratin and a nucleus containing DNA/RNA (Figure 7 Panel A). When the CTC undergoes programmed cell death or apoptosis, the membrane loses its structure and the cytoskeleton and nucleus begin to crumble (Figure 7 Panel B). Upon further degradation, TMP's are formed in which portions of the cells are enclosed within pieces of the cell membrane (Figure 7 Panel C).

TMP's will stain with fluorescently labeled monoclonal antibodies directed at the surface and intracellular targets such as, but not limited to, mitochondria, golgi apparatus, lysosomes, peroxisomes, actin, cytokeratins, CD41, CD61, CD 146 and tissue factor for examination by high resolution fluorescent microscopy. The present invention further considers characteristics of TMP's after fluorescent labeling with membrane dyes like DiOC12, electron microscopy of TMP' s, and their chemical composition using raman spectroscopy. For each of these analysis, size and density distribution of TMP's can be determined and assess the optimal labeling for identification of TMP and their discrimination from background in the blood of non-cancer patients. For example, clumped PE-cytokeratin antibodies may cause TMP background in the blood of non-cancer patients. The clumped PC-cytokeratin antibodies will not be enclosed by a membrane and could be discriminated by using a fluorescent dye binding to the membrane.

Because plasma has been routinely discarded in the Cellsearch system before enrichment with the EpCAM labeled ferrofluids, the complete TMP count may not be determined from a sample taken for only CTC enumeration. Plasma contains microparticles smaller than the TMP's found in the Cellsearch sample. The present invention considers the relationship between the larger and smaller particles. One method to address this is to stain blood from metastatic patients with fluorescently labeled antibodies directed against EpCAM, CDE45 nucleic acid and/or other membrane dyes and compare the two groups using flowcytometry or microscopic examination to provide information on their relationship.

A further embodiment of the present invention includes the apparatus for capturing and elucidating TMP. A 20 fold higher frequency of TMP's compared to CTCs was observed in a preliminary study. Therefore, blood volumes smaller than 1 ml will not be sufficient to yield the numbers of TMPs needed for patient application across all stages of cancers. Based on the optimal parameters for defining a TMP, a flow through device in which the input volume of blood can be varied to optimize the volume needed. While ferrofluids labeled with EpCAM for cell capture is one embodiment of the present invention, the present invention further considers capturing EpCAM positive events on a solid substrate followed by their release for analysis. An example is shown in Figure 8. Panels A, B, and C depict the making of a bead. A variety of chemistries can be used to functionalize the surface and couple streptavidin, for example, to the surface of the microspheres. In Panel B, the biotin sites of the streptavidin will be bound to desbiotin labeled EpCAM antibodies to create the functionalized beads, shown in Panel C. Panel D shows the bead capturing objects that express EpCAM. In Panel E, biotin is added at a higher affinity for streptavidin as compared to desbiotin which results in release of the desbiotin EpCAM antibodies and the EpCAM+ objects, shown in Panel F. A column with the beads in which EpCAM+ events can be captured is shown in Figure 9. Optimal packing of the microspheres and flow conditions for passing blood through the column with sufficient collisions to allow cells and TMP's expressing EpCAM to bind to the surface. All blood components that did not bind will be passed to the waste.

As an example, but not limiting, captured EpCAM+ objects in a range between 0.5μm and 5.0μm can be evaluated for CTC and TMP. A solution containing biotin will pass through the column (Figure 9, Panel B). The higher affinity of the biotin for the binding sites on streptavidin as compared with desbiotin will release the CTCs and TMP's expressing EpCAM. The fluid will then first be passed through a 5μm filter followed by passage through a 0.5μm filter (Figure 9, Panel C). After trapping the cells and TMP's both filters can be taken from the system and used for further examination (Figure 9, Panel D). Suitable filters for this application are produced by Aquamarijn MicroFiltration BV, Zutphen. The first filter with a pore size of 5μm contains approximately 0.25 x 10⁶ pores and the released cells are trapped on the filter. Using EpCAM capture from 7.5 ml of blood from healthy donors is between 500 and 5000 cells captured non-specifically. The excess of pores will avoid clogging of the filter. The second filter with a pore size of 0.5μm contains approximately 25 x 10⁶ pores and the released TMP's are expected to be trapped on this filter. The captured events on the filters are then fluorescently labeled and examined by fluorescent microscopy.

The present invention considers all fluorescent combinations for identification and separation from background. For example, positive identification candidates are fluorescently labeled streptavidin or fluorescently labeled EpCAM (different epitope), membrane dyes such as DiOC, or fluorescently labeled cytokeratins. Examples of candidates for negative identification include fluorescently labeled monoclonal antibodies that recognize leukocytes (CD45), platelets (CD41), erythrocytes (CD235a) and endothelial and platelet derived microparticles (CD146, tissue factor).

EXAMPLE 1

Circulating Tumor Cell definitions versus survival in Castration Resistant Prostate Cancer

Samples from patients enrolled into a prospective multi center clinical trial that evaluated the utility of counting circulating tumor cells for predicting response to therapy, progression- free survival, and overall survival in Castration Resistant Prostate Cancer (CRPC) patients were used for this study. Blood was collected before starting a new treatment and at monthly intervals prior to the next cycle of therapy. CTC were enumerated in 7.5 ml of blood using the CellSearch system (Veridex Raritan, NJ). A total of 276 patients were enrolled in the study. From this group 173 patients were selected that had CTC enumerated before initiation of a new line of therapy (baseline) and a follow up time point within 8 weeks after initiation of therapy (follow- up). In addition data from 6 patients were included that died within 10 weeks after baseline and before a follow-up blood sample could be obtained. Stored digital images from samples from 63 patients with only a baseline or follow-up sample were used to establish the CTC definitions and establish the ranking within each CTC class. Stored digital images from samples of 67 healthy individuals used in previous studies⁵ were reanalyzed to assess the background level for each of the CTC classes.

The CellSearch system (Veridex LLC, Raritan, NJ) consists of a CellTracks Autoprep (Veridex LLC, Raritan, NJ) for sample preparation and a CellTracks Analyzer II (Veridex LLC, Raritan, NJ) for sample analysis. The CellTracks Autoprep immunomagnetically enriches epithelial cells from 7.5mL of blood using ferrofluids coated with epithelial cell specific EpCAM antibodies and stain the CTC enriched samples with phycoerythrin conjugated antibodies directed against cytokeratins 8, 18 and 19, an allophycocyanin conjugated antibody to CD45 and the nuclear dye DAPI. The CellTracks Analyzer II is a four color semi-automated fluorescence microscope that captures digital images covering the entire surface of the cartridge for four different fluorescence filter cubes. From the captured images, a gallery of objects meeting pre- determined criteria is automatically presented in a browser for interpretation by a trained operator who makes the final selection of cells. Results of cell enumeration are expressed as the number of cells per 7.5mL of blood. In this study the stored images were used to explore the role of alternative CTC definitions.

Images from 484 samples that were stored on CD's or DVD's were successfully transferred to a hard disk (50Gb). Images were imported into the Linux software of the

CellTracks Analyzer II. The automated algorithm in this software was used to identify events in the cytokeratin (CK) and or DAPI images. The algorithm identifies events of at least 9 pixels in size of medium to high contrast in any selected channel. If more than one channel is selected for analysis events must be at least adjacent to each other before the algorithm presents the image to the user. Two sets of analysis were performed where in one set thumbnails of events are presented to the reviewer with staining of both CK and DAPI and in the other set events are presented only staining with CK. The operator reviews thumbnails of all events and is shown the maximum circumference of an event plus a boundary of at least 10 pixels around the event. Monochrome thumbnails show the staining in DAPI, CK, CD45 channels as well as a false color overlay of DAPI and CK to show degree of overlap between channels. The minimum size of the thumbnails is 40x40 pixels (25 μm²).

The reviewer of the images was blinded to both the survival of patients as well as the original CTC count. Once a total of 150 events were assigned to a class the result were extrapolated to the total number of events found.

CTC classes were defined using a training set of 63 samples that did not meet the inclusion criteria of having samples from both baseline and follow-up available. Statistical analysis was performed in SPSS 16.0. CTCs were subdivided into four groups of similar size Thirty-three baseline samples and 30 follow-up samples from the 63 samples were used to set the boundaries for each group on the 25 percentile, the median and the 75 percentile value. If there were more than 50% of samples with 0 events, the data was recoded into three groups. For all classes Kaplan Meier survival plots were generated and Cox regression was performed. Age was included as a continuous variable in all Cox regressions, with a typical CH ratio of 1.02/year with a CI of 1.002 to 1.04 and significance of 0.029 (WaId test). Race, processing site and the time between first diagnosis of tumor and baseline sample were not statistically significant contributors to survival and therefore not included in the Cox regression.

Results

A first set of four classes stained both with Cytokeratin (CK) and DAPI. The CTC classes were:

1. Intact CTC, round or ellipsoid nucleus (DAPI) entirely surrounded by a uniform CK stain at least 4 μm squared in size and not staining with CD45.

2. Granular CTC, any shape nucleus connected to CK with at least 3 higher intensity dots, at least 4 μm squared in size and not staining with CD45.

3. Large Tumor Cell Fragment (L-TCF), a nucleus of at least 4 μm squared with any size CK and not staining with CD45.

4. Small Tumor Cell Fragment (S-TCF), a nucleus smaller than 4 μm squared with any size CK and not staining with CD45. Typical examples of these four classes are presented in the top Panels of Figure 10. Within one review an event was only assigned to one class or not assigned at all.

A second set of four classes stained with cytokeratin only. The classes were:

1. Large Tumor Micro Particle (L-TMP), CK staining area larger than 4μm square, DAPI + or DAPI-.

2. Small Tumor Micro Particle (S-TMP), CK staining area smaller than 4μm square DAPI + or DAPI-.

3. CK and CD45 positive fragment (CK+/CD45+), any DAPI + or DAPI- fragment that has CK and CD45 signal.

4. Focus Artifact (FA), CK staining that looks like a ring with a single dot in the middle, similar in appearance to an out of focus bead. Clusters of rings were also considered FA.

Typical examples of these four classes are presented in the bottom Panels of Figure 10.

All CTC classes were prognostic for survival (log rank p <0.001) both at baseline and follow- up except for the CK+/CD45+ class and FA class that were used as control classes. Survival prospects decreased with increasing events for all relevant CTC classes as evidenced by the shortest survival for group 4 (G4) followed by G3, G2 and then Gl. This is illustrated in Figure 11 showing the Kaplan Meier plots from follow-up samples using the CellSearch CTC definition in Panel A, the CTC class Intact CTC in Panel B, Granular CTC in Panel C, L-TCF in Panel D, S-TCF in Panel E, L-TMP in Panel F, S-TMP in Panel G, CK+/CD45+ events in Panel H and in FA events in Panel I. No separation of the curves is noted for the CK+/CD45+ events (p=0.36) and FA events (p=0.73) whereas all other CTC classes show a clear separation.

Multicenter prospective clinical studies demonstrated a significant relation between the presence of CTC and poor progression free and overall survival. The definition of the objects classified as CTC were set before the initiation of these studies and no analysis has been performed to determine what the best CTC definition is to predict outcome. The rapidly increasing number of reports on CTC identified with various technologies urges the need for guidelines for defining a CTC. In this study we restrict ourselves to CTC identified by the CellSearch system that needs -2000 EpCAM antigens and -20.000 Cytokeratin 8, 18 and or 19 antigens on a CTC for it to be presented to a reviewer as a CTC candidate.

Stored digital images from a prospective multicenter CRPC study were then reanalyzed using different criteria based on size and shape of the nucleus and cytoplasm of the CTC. From the 8 defined classes of CTC, two did not require the presence of nuclear material and were named large tumor microparticles (L-TMP) or small tumor microparticles (S-TMP). Surprisingly the correlation with survival was stronger for these two classes compared to intact CTC. One explanation is that the frequency of TMP which are an order of magnitude larger resulting in a more sensitive and accurate assessment. The presence of TMP thus could infer that intact CTC are present albeit at a frequency below 1 in 7.5 ml of blood. Cleavage of cytokeratin results in a granular appearance and granular CTC class may be associated with CTC undergoing apoptosis. Further, the relative proportion of "Granular CTC" increases after initiation of therapy, suggesting an effective therapy. Relation with outcome I this study was similar between intact and granular CTC suggesting that for this application appearance is not important for prognostic value. From the survival analysis presented here, very different morphological CTC definitions are useful for clinically relevant information.

Both the Kaplan Meier plots and Cox Hazard ratios show that for most CTC classes, a higher number of events is associated with a shorter survival. The "Intact CTC" and "Granular CTC" are the only classes where a dichotomization might be of similar value this however is likely due to the low overall densities of these classes. The ranking within each subclass was based on an independent set of 63 samples to avoid bias in the determination and, thus, provide a means for future investigations to use this classification in practice.

The TMP are easy to count, while CellSearch CTC L/S-TCF require the most operator training. The present study shows that a fully automated system to identify CTC/TMP should eliminate the variability in the assignment of objects. CTC/TMP is more easily standardized to develop a fully automated image analysis.

Although the present invention has been described with reference to specific

embodiments, workers skilled in the art will recognize that many variations may be made thereform, for example in the particular experimental conditions for TMP particle separation herein described, and it is to be understood and appreciated that the disclosures in accordance with the invention show only some preferred embodiments and advantages of the invention without departing from the broader scope and spirit of the invention. It is to be understood and appreciated that these discoveries in accordance with this invention are only those which are illustrated of the many additional potential applications that may be envisioned by one of ordinary skill in the art, and thus are not in any way intended to be limiting of the invention. Accordingly, other objects and advantages of the invention will be apparent to those skilled in the art from the detained description together with the claims.

Claims

We claim:

1. A method for diagnosing disease in a test subject comprising:

a. obtaining a biological specimen from a test subject, said subject suspected of containing disseminated tumor cells and tumor cell related microparticles (TMP); b. preparing a magnetically- labeled sample wherein said biological sample is mixed with magnetic particles coupled to a first biospecific ligand which reacts specifically with said disseminated cells, and said TMP, to the substantial exclusion of other specimen components;

c. contacting said magnetically-labeled sample with at least one additional biospecific ligand which specifically labels said disseminated cells and said TMP to the substantial exclusion of other specimen components;

d. analyzing said labeled disseminated cells, and said labeled TMP, the presence of said labeled disseminated cells, said labeled TMP indicating the presence of disease.

2. The method of claim 1 wherein the biological specimen is patient blood.

3. The method of claim 1 wherein said first biospecific ligand is EpCAM.

4. The method of claim 1 wherein said TMP is cytokeratin positive, CD 45 negative, and less than 4 micrometers.

5. The method of claim 4 wherein said cytokeratin positive is determined by antibodies directed against cytokeratins 8, 18 and 19.

6. The method of claiml wherein said analyzing is the CellSearch System.

7. The method of claim 1 wherein said disseminated cells are cytokeritin positive, CD45 negative and positive for nuclear dye stain.

8. A method for diagnosing disease in a test subject comprising: a. obtaining a biological specimen from a test subject, said subject suspected of containing tumor cell related microparticles (TMP);

b. preparing a magnetically- labeled sample wherein said biological sample is mixed with magnetic particles coupled to a first biospecific ligand which reacts specifically with said TMP, to the substantial exclusion of other specimen components;

c. contacting said magnetically-labeled sample with at least one additional

biospecific ligand which specifically labels said TMP to the substantial exclusion of other specimen components;

d. analyzing said labeled TMP, the presence of said labeled TMP indicating the presence of disease.

9. The method of claim 8 wherein the biological specimen is patient blood.

10. The method of claim 8 wherein said first biospecific ligand is EpCAM.

11. The method of claim 8 wherein said TMP is cytokeratin positive, CD 45 negative, and less than 4 micrometers.

12. The method of claim 11 wherein said cytokeratin positive is determined by antibodies directed against cytokeratins 8, 18 and 19.

13. The method of claim 8 wherein said analyzing is the CellSearch System.