Delivering on the Promise of Liquid Biopsy

Learning Objectives

The learning objectives for today are to understand how rare cell analysis, including liquid biopsy, can be incorporated into diagnostic pathology to extract clinically important information, as well as to understand the value of phenotypic and molecular characterization of individual rare cells within liquid biopsies and other specimen types.

Evolution of Tumor Sampling

The approach to accessing tumors for analysis has evolved from highly invasive to increasingly less invasive. Yet at the same time that biopsy sample volume has decreased, the diagnostic information required from the sample has increased. With liquid biopsy, we have a non-invasive sampling method with very few cells. The challenge, then, is to extract information from these few cells.

History of Liquid Biopsy

First, a little bit of the past of liquid biopsy. The term "liquid biopsy" has quite rapidly entered the vocabulary of diagnostics. Its original meaning was to describe a way to access the cells of a tumor indirectly by capturing rare tumor cells that circulate in the blood and thus allow a real-time investigation of the cancer via blood sampling.

Before discussing the future, let's actually go even further back to its original origins. As many of you know, the origins of the field can be traced to nearly a century and a half ago in the famous report from Thomas Ashworth in the Australian Medical Journal. In this paper, he described a patient who had many tumors visible in the skin throughout the body. He investigated the blood to try to understand how so many tumors appeared. The tumors had cells with an unusual appearance and Ashworth saw similar cells in the blood, allowing him to infer a relationship between the tumor cells and the cells he found in the blood.

Here are the figures that he drew of the tissue and of the blood cells.

Below is a micrograph of a chordoma, which is likely what the tumor that Ashworth described as a "tumor of chorda dorsalis" in his paper actually was. You can see a similarity between the cells in the micrograph and in the drawings. Microscopic visualization of the cells is what actually allowed Ashworth to make his investigation and arrive at his seminal conclusion. That was the past.

First Generation Liquid Biopsy

The present has largely been influenced by the CellSearch® system. This can be considered the first generation of liquid biopsy. CellSearch® accomplished remarkable milestones for liquid biopsy, notably the prognostic significance of CTC counts in epithelial cancers.

Next Generation Liquid Biopsy

The future of liquid biopsy will involve expanding the breadth of diagnostic analyses to match what can be obtained with tissue biopsy. CellSearch® has robustly addressed the first level of tissue biopsy evaluation, namely identification and counting of circulating tumor cells.

A second level of evaluation is phenotypic characterization of malignant cells, as well as potentially immune cells in the tumor microenvironment.

A third level is the molecular analysis of individual cells. Here is an example of what next generation liquid biopsy will look like. These are unusual cells that were found in the blood of young woman. You can see in green that the cells are expressing cytokeratin. This woman had no history of cancer. She did, however, have a history of pregnancy. These are in fact fetal cells circulating in maternal blood. We know this because we isolated them individually and demonstrated on a single cell basis that they were male cells in a female host. After removal of individual cells, whole genome amplification, and CGH array analysis, we can see in the left panel the presence of a Y chromosome in the cell in a maternal sample.

Similar analysis of blood from a woman known to be carrying a fetus with trisomy 18 confirmed the presence of the additional copy of the chromosome by both CGH array and next generation sequencing, as you see in the panel on the right.

Even more remarkably, a 2.7 megabase micro-deletion was confirmed in a circulating fetal cell in another individual. This demonstrates the feasibility of using fetal cell analysis to eventually replace amniocentesis or chorionic villus sampling for testing for genetic abnormalities. In effect, this is a liquid biopsy of the placenta, since these cells have trophoblast origin.

As we know, traditional tissue biopsy collects tumor samples by an invasive procedure, processes the tissue, typically using formalin fixation and paraffin embedding, to a slide. Sections are then stained to identify the cancer and further characterize it by type [phonetic], C [phonetic] phenotyping, and biomolecules from the tissue sections may be isolated to perform molecular analyses.

At RareCyte, we have developed a platform that is in many ways a direct descendant of Ashworth's analysis 150 years ago. It is modeled after established diagnostic pathology workflow to put cells on a slide. Once on a slide, what can be done with a tissue section can now be done theoretically with rare cells. Cells may be identified, characterized by phenotype, and then analyzed at a molecular level.

AccuCyte® Sample Preparation

The first step in the process is called AccuCyte®. It is sample preparation. AccuCyte® separates and collects the nucleated fraction of the blood that contains the rare cells of interest within the buffy coat. It is a user-friendly system that exploits the characteristic density of nucleated cells, which are transferred then to a microscope slide. Since the buffy coat is collected in the process without washes, the plasma can also be collected from the same blood sample. Once on the slide, the cells can be stored on that slide in the freezer for a long period of time. The prepared slides can then be placed into automated immunostaining instruments for immunofluorescence staining with multiple markers. The standard markers that we use to identify CTCs include epithelial cell markers (Cytokeratin, EpCAM), a leukocyte exclusion marker (CD45), and then a nuclear stain. The use of automated strainers eliminates staining variability and increases control of conditions to allow for amplification of low expression targets, and as we will discuss more later, up to six markers to be simultaneously assessed.

CyteFinder® Automated Scanning

Stained slides are then imaged with RareCyte's CyteFinder® automated fluorescence microscope. CyteFinder® is a precision-engineered instrument that scans a complete slide in about 11 minutes and has up to 6-channel imaging capability. The scanned image is analyzed by an image analysis algorithm in about 10 minutes. The algorithm identifies candidate target cells that have the characteristics of the cells of interest and presents them to a reviewer for confirmation. The algorithm ranks candidate cells based on the defined parameters that determine how likely a cell is to be a true CTC and gives it a score between zero and 100. The highest scoring cells are presented to the reviewer first. This increases consistency of reviews, decreases user or observer fatigue, and also speeds the whole process and makes it more efficient.

CytePicker® Single Cell Retrieval

Integrated within the CyteFinder® instrument is the CytePicker®, a built-in module that mechanically dislodges individual cells and pulls them into the needle tip of the instrument. Cells can then be expelled into an imaging tube, allowing visual confirmation of the individually retrieved cells before single cell molecular analysis. The CytePicker® is highly automated and is designed so it does not require a high degree of technical skill.

Here are some examples of lung and breast cancer circulating tumor cells imaged using the 40x objective, also part of the CyteFinder® [phonetic] instrument. You can see that there are many cytological details visible in the cells.

CTC Count Monitoring Over Clinical Course

Here is a case study that used the RareCyte® platform for the monitoring of CTCs as well as single cell isolation for genomic analysis. In an innovative clinical study directed by Tony Blau at the University of Washington, nearly 50 CTC analyses over nine months were performed on blood from a patient with triple-negative breast cancer. Her CTC count paralleled her clinical course. Her baseline of more than 1,000 CTCs per mL dropped 100-fold during platinum therapy, and then climbed again as she progressed. She did not respond to single agent targeted therapy with crizotinib, and that was chosen because the patient had a ROS1 mutation that was confirmed in circulating tumor cells, that was theoretically a potential driver of the cancer. She had a brief response to eribulin at the end of her course that was also reflected in a reduction of her CTC count once more.

Surprisingly, when she died, she had a lower CTC count than she had had two months previously. This may be because before she died, she had very large clusters of circulating tumor cells in her blood. This is one of the most dramatic and notable clusters that we found. You can see that there is tremendous heterogeneity of size and shape as well as heterogeneity of EpCAM phenotype. EpCAM is marked in red. You can see some cells are no larger than a lymphocyte and are cytokeratin positive but EpCAM negative. Others are very large with very prominent EpCAM staining.

The cause of this patient's death was respiratory failure. Here is a microscopic image of her lung at autopsy. Although the alveolar air sacs are open, the arterials are clogged with tumor cells, antalye [phonetic] that kept blood from circulating through the lung. This was a case in which a patient died quite literally from circulating tumor cells present in her blood.

Single CTC Genomic Evolution in Context of Therapy

One of the advantages of the archivability of the samples is that we can go back to samples that were collected earlier and perform molecular analyses. We retrospectively isolated CTCs for single cell genomic analysis from six time points during the patient's course. Individual CTCs were picked with CytePicker®, single genomes amplified, and whole exome sequencing performed with our colleagues Anup Madan and Kellie Howard at Covance Genomics. Multiple bioinformatic filters were applied to generate a list of sequence variants for each cell.

Identification of Mutations in Individual Cancer Cells

The results are shown here in this figure. Each column represents a single CTC grouped by time point. Each black bar represents a mutation that was identified in that CTC. The mutations can be classified into three groups.

First, a static group that remains unchanged during the course of therapy. You can see that group at the top of the figure, across all time points. Sample-specific groups that are unique to each time point, so for instance you can see at Day 6, at the bottom of the group of mutations, there are variants present that are not present at Day 91. Then, in addition, there are intermediate groups that have variants that are short-lived but are present across multiple time points.

Note that the cells on Day 91 are not directly derived from cells on Day 6 since they do not include many of the mutations that are found on Day 6 cells. This suggests that they are different branches from the same clonal [phonetic] tree, related at the trunk level but not directly springing from the previous set of CTCs that were evaluated.

Mutations in Context of Clinical Response

We further investigated the variants to determine how many were predicted to be deleterious, or were known reported cancer drivers. Cancer driver mutations increased after exposure to therapies that produced clinical response. You can see that at Day 91 relative to Day 6. Over time, you could see that as the patient began to progress, the number of cancer drivers decreased. Then after the successful, brief treatment with eribulin, with CTC count decreasing, the cancer drivers again at Day 258 came up. Interpretations of these data are really quite intriguing. One possibility is that the cells with increased cancer driver mutations represent malignant cells that are resistant to the current therapy.

Multi-Parameter Phenotyping

RareCyte has built a framework for custom assay development that utilizes the multi-channel capability of the platform. We call this the "4+" framework. Four fluorescent channels are used for the identification of canonical CTCs, using stains for nucleus, the epithelial markers cytokeratin and EpCAM, and the white blood cell exclusion marker CD45. This leaves two additional channels available for biomarkers of specific interest. High amplification is possible for targets that may be expressed at low protein levels. In fact, if cytokeratin and EpCAM are combined into one epithelial channel, then three additional biomarker channels are available for investigation, further extending the depth of characterization available for their [phonetic] cells.

The following slides demonstrate how the 4+ framework allows biomarker investigation in prostate cancer.

Prostate "4 + 1" - Androgen Receptor (AR)

Here is a 5-channel assay that adds androgen receptor or AR to the four channels. The androgen receptor is a principal driver of prostate cancer. Here you can see that its staining location is predominantly nuclear and this cell line spiked into blood.

Prostate "4 + 1" - ARv7 Variant

The androgen splice variant ARv7 has been associated with resistance to endocrine therapies such as enzalutamide and abiraterone. Here you can see its presence also in a nuclear location.

Prostate "4 + 2" - AR + PSA

This image is of a 6-channel assay in which prostate-specific andogen and andogen receptor have been added into channels five and six. The other markers are not shown in this particular image.

PD-L1-positive CTCs in Breast Cancer

Here is a 6-parameter assay applied to a sample in breast cancer. In this assay, the immune checkpoint regulator PDL1 is identified in the fifth channel, and the interferon regulatory factor 1, an interferon-gamma-induced transcription factor that regulates expression of PDL1, is in channel six. Recent studies comparing PDL1 and IRF1 tissue staining have shown that IRF1 had a higher predictive value than PDL1 in determining response to checkpoint inhibitors against PD1 or PDL1. You can see here that the localization of IRF1 is in the nucleus of cells in a CTC cluster.

CTC Drug Target Companion Diagnostics

The promise of liquid biopsy really is to make possible via blood sampling what was previously available only via tissue biopsy. This includes companion diagnostics. Here is an example of a drug target marker that RareCyte has developed with a pharma partner.

Here you can see distinctly in green the presence of the drug target on a cluster of cells that were collected from a patient-derived xenograft model and spiked into human blood.

5-cell CTC Cluster Expressing Drug Target

The same assay was subsequently applied to blood from cancer patients. Here we see a cluster of cells that express the target of the drug under study. This assay is now being implemented in clinical trials of the drug.

Rare Cell Samples in Diagnostic Pathology

The technology that RareCyte has developed for identifying circulating tumor cells in liquid biopsies may also be rationally applied to other sample types in which rare cells are present. Often, a definitive diagnosis is not possible in small volume cytology samples, such as fine needle aspirates, and additional biopsies are often required.

Identification and Retrieval of Malignant Cells in Hodgkin Disease FNA Sample

Here is an example of a fine needle aspirate sample of the lymph node in a patient with classical Hodgkin lymphoma that we evaluated in collaboration with Jonathan Fromme and David Wu at the University of Washington. Using a 6-parameter panel, cells with a phenotype that included CD15, CD40, and CD30 were identified using CyteFinder® and then picked individually using CytePicker®.

Molecular Analysis of Single Cells from Hodgkin Disease FNA Sample

Analysis of the heavy chain of the B-cell receptor identified PCR products with precisely the same length from these two cells, providing very strong molecular evidence that the cells were in fact from the same clone, thus confirming that the cells identified were in fact likely to be Hodgkin cells.

Pancreatic Cancer Tissue Section Cell Retrieval

The RareCyte® platform can also image tissues by immunofluorescence. Individual regions of cells can also be picked by CytePicker®. Methods for molecular analysis of cells picked from tissue are under development. This will allow the correlation between phenotype, including morphology [phonetic], of cancer cells and their genotype. Furthermore, molecular analysis of atypical cells could be performed to determine whether or not they are in fact malignant.

Minimal Residual Disease Monitoring: Multiple Myeloma

Finally, the platform technology that was designed to identify cells from solid tumors in a liquid biopsy may also be used to identify minimal residual disease in liquid tumors. Here is an example of CD138 positive malignant cells identified in the blood of a man with multiple myeloma.

Making Non-Diagnostic Samples Diagnostic

I envision that in the future, samples that are initially non-diagnostic due to insufficient numbers of atypical cells, could undergo rare cell picking and molecular analysis. In many of them, molecular alterations that indicate presence of cancer could effectively turn them into diagnostic samples.

It has been a pleasure to speak to you about the promise of liquid biopsy and how methods developed for it can be applied to other sample types. Thank you very much for your attention. I would be pleased to answer your questions.

Q & A

Q: Is liquid biopsy useful for mesenchymal tumor or only for carcinomas?

Dr. Kaldjian: That is a very good question. Let me start by giving a little bit of background first. Our technology allows the identification of any rare cell that is present in the blood as long as there are markers for that rare cell that can differentiate it from the other cells that are present. So if one were to, for instance, look for circulating melanoma cells, one would use a panel of markers that included melanoma-specific markers. That is quite possible and we have developed such prototype assays.

In addition, if one were to look for, say, circulating sarcoma cells, again, the challenge there would be to come up with a list of antibodies that would differentiate your sarcoma from other cells that were circulating in the blood. With the great interest in epithelial cancers that undergo the epithelial to mesenchymal transition, we have developed a combined epithelial-mesenchymal assay, which includes vimentin, in order to be able to identify such cells.

Q: Is there a current process to dissociate solid tumor cells to allow the use of this technology?

Dr. Kaldjian: RareCyte has not embarked on that type of evaluation, which would be in a sense analogous to what is done by flow cytometry. Rather, there are some users that are doing some rather novel work in the tissue area to be able to preserve tissue architecture and still identify rare cells. So in this situation, one would use fluorescence staining of the tissue, and then using the same type of algorithmic approach to identify rare cells, identify those cells within the tissue, and then pick those cells. The caveat there is when cells are adjacent to other cells it can be very challenging to get one cell. Rather, it is quite possible to get a few cells there. So it is not to say that this is not possible. I just would say that it has not been the focus of our efforts. But it would certainly be something that could be done. One could make smears from the disaggregated tumors and then try it this way.

Q: What are the advantages of cell-based liquid biopsy over cell-free DNA analysis?

Dr. Kaldjian: That is a very timely question, too. As many of you know, the general term liquid biopsy is more tightly associated with cell-free DNA analysis currently than with cell-based analysis. I think that a few years ago, the field generally thought that since the cell-free approach and evaluation of the plasma seemed so straightforward, this would really be the future of genotyping. The approaches, I believe, and I think many in the field believe, will be complementary. They are different tools, and depending on the situation, a different tool should be applied.

Oftentimes, one hears this analogy of the difference between a fruit smoothie and a fruit salad. The idea being that cell-free DNA analysis gives the sum total of all the fruit that has been ground up in that fruit smoothie. One can identify mutations but not necessarily know what fruit they came from. Whereas, the fruit salad approach is being able to get individual cells, and now if there are different mutations identified, one can ask, are they all present in the same cell or are they in different cells?

I think another potential advantage is that the DNA there is pure. There is no background DNA that requires bioinformatic analysis or any other method to separate. So I would say that there are going to be complementary situations in which one or the other will be, I think, potentially preferred, much like it is in the fetal cell analysis field right now.

Q: Now four plus two means that any additional two markers can be added, is that right?

Dr. Kaldjian: Yes, what that means is additional markers can be added. Currently, there are minimal constraints regarding species. So generally speaking, if your antibodies are standard rabbit or mouse species then they should work quite nicely in the additional two channels. The assays, of course, require validation, similar to any other type of immuno-based assay validation.

Q: How small in size and what stage tumor can be identified with liquid biopsy, in your practice, for example?

Dr. Kaldjian: That is another really good question. If you scan the literature, you will find that if you look at the whole spectrum of cancers from stage I to stage IV, epithelial cancers, you will find the highest incidence of CTCs that have been identified in the stage III and the stage IV categories. Not that they do not exist in stage I or stage II, but they traditionally have been a lot harder to find. And I think the reason for not finding so many can be related to many different factors. One could be the overall sensitivity of the assay. Another reason could be due to the actual volume requirements. If there is only one CTC that is present in, say, 20 mL of blood, but the typical sampling is 7.5 mL of blood, you can see it is a statistical question of how likely it is to be able to identify that cell. One way I think about the progression of cancer is that all stage IV cancers were at one time stage I cancers. All metastatic tumors were once CTCs. By this I mean distant metastases. So the sense that I have based on our own experience is that as the methods become increasingly sensitive, we are going to find more and more evidence further up the cancer progression scale, or stage, in which it will be very rational to apply these methods, perhaps screening in high risk populations, and certainly the monitoring of recurrence after treatment.

So even though my answer is not definitive and it does certainly depend on cancer type, I think that what we are going to find is that in general, the later the stage, the higher the incidence of CTCs being identified, but increasingly we are going to find more and more situations where lower stage tumors have CTCs that could be meaningfully evaluated.

Q: When do you think that cell-based liquid biopsy will enter clinical pathology practice?

Dr. Kaldjian: That is another very interesting thing to think about. As a young biotechnology company, our approach is to develop the technology initially for research use and then the next likely step is going to be to develop it for laboratory test use, via CLIA laboratories, primarily in early-adopting academic centers. Then eventually reach that point, very likely via companion diagnostics, where we get regulatory approvals at the federal level. So I think there is probably a pretty good chance that within as early as the next year there will be initial academic adoption in CLIA laboratories of the use of the RareCyte® platform. Then I suspect that would quickly lead to additional applications within the CLIA setting. Once there is a formal approval of a test at the FDA level then I think it will become much, much more rapid.

The other possibility, too, is that if in fact we can apply these methods effectively to make non-diagnostic cytology samples into diagnostic ones, I can see the entrance of this technology into the standard pathology workflow very rapidly.

Q: Is there a current process to disassociate solid tumor cells to allow use of this technology?

Dr. Kaldjian: When it comes to the actual tissue staining, there is a method that is currently being developed by one of our academic collaborators in which the tissue is stained with one round of immunofluorescence staining, the tissue is imaged, and then the stain is in a sense wiped clear from that slide. Then another round of staining is performed with a new set of markers, and then that is imaged. And this is repeated for several to many rounds. What that in effect does is generate deep multi-dimensional information in which that tissue can be interrogated at a single cell level. And I think this is going to become a very, very interesting research tool to help understand the heterogeneity of tumor cells as well as tumor microenvironment cells, including tumor-specific immune [phonetic] cells.

Q: Would it be possible to exfoliate cells from a fresh tumor and achieve a cytology [phonetic] in order to use this technique?

Dr. Kaldjian: If by exfoliating cells you mean something like making a touch preparation or some other method in which you have free individual cells or clumps of cells in fluid, yes, that is quite possible. We have actually used versions of techniques such as ThinPrep [phonetic], as well as our own method, to be able to take cells in suspension, spread them onto slides, and then evaluate them in this way. And continued efforts in this arena are actually ongoing.

Q: Can other cytology preparations besides fine needle aspirates be analyzed with this platform?

Dr. Kaldjian: Yes. That actually follows on the previous question, in which we just touched on touch preps. We have done some initial proof-of-concept work, again with academics, looking at urine specimens and then processing urine specimens using the ThinPrep method onto slides. Then being able to stain with our standard four marker panel, identify the presence of spiked in cells into human urine, and then be able to pick them and perform molecular analyses and demonstrate that, yes, in fact we could identify or confirm mutations present there. So in general sense, you can consider rare cell problems as being a few cells in the background of many, many cells. Or another way to think about it is a few cells that are present in a very large volume. Urine is such a specimen type; cerebral spinal fluid would be another. Certainly, pleural [phonetic] fluids and ascites fluids could be other types of fluids that would fit there. And frankly, I see no reason things like bronchial brushings, and so on, could not also fit within this paradigm.

Q: Can other rare cells besides cancer and fetal cells be identified with the RareCyte® platform?

Dr. Kaldjian: I had mentioned a couple of times briefly the possibility of looking at rare immune cell phenotypes. I did not show slides here, but once one has six marker channels available, one can start to subtype lymphocytes according to functional categories. By using such markers, such as CD3, CD4, CD8, and then other markers such as PD1, Tim--it is either Tim-1 or Tim-3, I am blanking on it--but the exhaustion markers, one can find T cells of interest that are present at very, very low concentration within the blood.

In addition, we have done some really fascinating work with other colleagues at the University of Washington identifying antigen-specific T cells, both CD4 and CD8, which react against a polyomavirus that is the causative agent of Merkel cell carcinoma of the skin. So I think this reinforces the concept that if there is a rare cell that is present in a sample, and if there are markers that can differentiate that from other cells in the background, then we should be able to find it, whether or not it is a tumor cell.

Q: Is it possible to stain for more than six markers?

Dr. Kaldjian: As we discussed just recently, this ability to be able to stain, and then to unstain, and then restain, allows that possibility to stain for multiple markers and get the deeper level of information. At this point, we can only stain for six markers simultaneously, although I think our engineers are looking to see whether there are ways to be able to even extend that to additional channels.

Q: Would you say that with the possibility of doing a touch prep on fresh tumors this technique comes as an aid and even sort of a substitute to frozen sections?

Dr. Kaldjian: I'm thinking on the fly right now. I have never actually considered how this could be used as a substitute in real time for frozen sections. But I think this might be challenging, and the reason for that is as the workflow is currently set up, the amount of time it would take to stain on the automated stainer using our standard approach would probably be too long to be able to keep the patient in the operating room, to be able to generate that result. It is an interesting idea, and I think what you would have to do is come up with a more minimal set of markers and maybe another approach that would allow the staining to be done very, very quickly. But right now, I think the major constraint would be the amount of time required in order to be able to get the surgeons the answer that they need. Currently, a typical run on the automated stainer will take, kind of like the regular IC [phonetic] run, anywhere from 1-1/2 to 2-1/2 or 3 hours. So I think that would be an impossibility right there. But it is an interesting idea to think about, and I appreciate the question.