Akoya Biosciences offers 50+ protein markers along with existing Phenoptics offering, while Berkeley Lights has a unique single-cell, live-cell analysis platform
Every year the American Association for Cancer Research conference is a major conference for major product announcements and new companies to test the cancer research market. And while walking around the exhibit floor, you can see which poster topics are crowded. For example, Section 17, “Circulating and cell-free biomarkers for diagnostics and monitoring of cancer 1” and Section 18 “Current developments in non-invasive biomarkers for the assessment of cancer 1” were two areas that you could hardly walk through, there were so many people in these isles (yours truly among them).
Another section, Section 20 “Imaging the Tumor Microenvironment” had this poster from Akoya called “Highly multiplexed single-cell spatial analysis of tissue specimens using CODEX” and you can access their abstract here.
Akoya Biosciences also had an exhibit-floor presentation this year, with Dr. David Rimm of Yale University as a guest speaker.
Akoya Biosciences CO-detection by inDEXing System
Akoya describes itself as the leader in ‘quantitative tissue biomarker evaluation’, which perhaps needs some explanation. In current practice with a immunohistochemistry, a particular tissue may be deemed ‘PD-L1 positive’ should a specific antibody (okay, say a Agilent Dako 22C3 PharmDx one where the PMA approval statement can be found here) light up a particular cancer tissue sample in a clinical setting.
This is qualitative, not a quantitative test: a Tumor Proportion Score of >=1% is ‘positive’, while a Tumor Proportion Score of >=50% is ‘high positive’, where the TPS is the fraction of tumor cells that stain with this particular antibody.
A quantitative test is a different animal: here you are looking at fluorescence signal intensity, comparing it against background, and taking quantitative measurements is a virtual measurement of concentration of protein in a field of view, a quantitative measurement of signal density.
Dr. Julia Kennedy-Darling gave an overview of the problem to solve: the complexities of the Tumor MicroEnvironment, commonly called TME, where there are many different types of cells, many different kinds of influences, starting with the extracellular matrix (ECM), different signaling molecules from different immune celltypes, cancer-associated fibroblasts (CAF), dendritic cells, NK cells, B cells, T-helper cells, T-killer cells, adipocytes, the list goes on. She then showed all the immune checkpoint markers (different CD molecules) that are known and commonly tested for – a list of 39 of them.
There is a need for a new system, and the CODEX system (CO-detection by inDEXing) is a fluidics system that works in conjunction with three different fluorescent microscope system software (Zeiss and Leica and one other were mentioned). A device holds a coverslip with the tissue slice on it on the standard microscope stage, and a gasket and top slip make a fluidics chamber that the Akoya instrument controls.
Their system works with different primary antibodies that are conjugated with oligonucleotides, and a set of fluor-labeled oligos (each oligo labeled with one of four fluors). The tissue is stained with the entire collection of up to 50 different antibodies at once in a single-step; then collections of three or four different identifier oligos with three or four different fluor dyes (Dr. Kennedy-Darling indicated the use of Alexa 488, Atto 550, and Cy5 in her initial presentation, with an option to use Alexa 750 if I have my notes correct). Cycles of imaging, stripping of only the labeled oligo (presumably by melting off the oligo hybridization via heat or chemically), another set of three or four fluor-labeled oligos hybridized and imaged, and the cycle repeats up to 16 times.
Dr. Kennedy-Darling spent some time in talking about signal and noise ratios, and how their experiments carefully controlled for different kinds of variables to insure repeatability and robustness. She also showed data comparing cycle 1 to cycle 11 without any loss of signal, which makes sense as only the oligos are removed, not the antibodies.
Next she discussed the process by which they validate different antibodies, the kinds of saturation and titration experiments they perform on an antibody-by-antibody basis, and how this information is contained in the Datasheet that comes with the specific antibody. She then showed clinical sample data from both breast cancer lymph node as well as renal cell carcinoma, and an interesting representation of the 29-marker single-cell data sets in 2-dimensional space exported in something called a .fcs file that retains both the X-Y coordinates as well as the 29-marker data, and represents them phenotypically using a tool called Vortex.
This 2016 Nature Methods paper, titled “Automated mapping of phenotype space with single-cell data” gives the details of this Vortex software, and the Github repo for Vortex is located here. Vortex enables the representation of cell proximity by cell-type, demonstrated in the Voronoi plot below.
Phenoptics Technology by Dr. David Rimm (Yale University)
As the CODEX system was only recently launched, Dr. David Rimm (Director of Pathology Tissue Services among other roles at Yale University) presented data on the recently acquired (from Perkin Elmer) Phenoptics technology which uses a different scanner called Polaris and a reagent system called Opal. The Phenoptics technology has two other iterations, called Vectra and Mantra, still on the Perkin Elmer website.
Dr. Rimm shared accepted-for-publication but still under embargo data with an interesting meta-analysis study that compares different assay biomarker modalities for PD-L1, and derive AUC curves from the data that can be analyzed together.
He showed a busy diagram that laid out their original number of studies (totaling 243), then 199 passing one set of criteria and after additional criteria ended up with 44 studies. These 44 studies had a total of 55 analyses associated with them, involving in total 8021 samples.
The ‘payoff’ slide is below, with the conclusion “The AUC of mIHC/IF than the other AUCs by both approaches”, including PD-L1 IHC, Tumor Mutational Burden (TMB DNA), Gene Expression Profiling (GEP RNA), where mIHC/IF is multiplex Immunohistochemistry / Immunofluorescence. One criticism of the mIHC/IF data is the lower number of samples with this technique, which he acknowledged.
Berkeley Lights single-cell technology
For those unfamiliar with Berkeley Lights, I previewed a poster a few years ago at the Advances in Genome Biology and Technology Conference where Dr. Bobby Sebra from Mt. Sinai showed this very interesting single-cell technology platform. (You can access that post here.)
Here at #AACR19 I spent some time with Dr. Mark White their Director of Scientific Affairs, who showed me the Beacon chip and several videos of how it works. Each ‘pen’ is a chamber that can control single-cell partitioning, and each chip (pictured below) has some 3500 ‘pens’ in them.
By ‘pen’, these are 1 nL (that is nanoliter) wells and the key benefit is that a particular dimension of an individual cell, such as cell proliferation as an example, can be individually targeted and harvested for other downstream uses. Thus a given cell population while considered ‘monoclonal’ (i.e. cells that had started as a single cell and then grown to a relatively large number) will still have a distribution in individual attributes. In biopharma, producing a monoclonal antibody hybridoma or an engineered CHO line (Chinese Hamster Ovary cell line, a mainstay in biopharma for those not familiar) can be optimized by this system for the individual cells that grow faster or produce more as a phenotype.
And out of 3500 individual cells on one of their OptoSelect™ Chips, the user can select the one that has the top antibody production or best growth characteristic to harvest and move downstream to scale-up and further testing.
The system works with each chip containing 4 million photosensitive elements, that when illuminated will create a dieletric field that can manipulate single cells. Thus through image recognition and manipulation of light individual cells (or collection of cells in an individual pen) can be segregated or expelled for collection.
This first video (kindly provided as an in-booth demonstration) shows the machine-learning pattern recognition to pick out individual cells and put them into individual pens.
This second video shows a simple proliferation time-lapse. Given the doubling rate of these cells may range from 10 hours to 36 hours, this video obviously represents several days of incubation.
This third video shows how an individual pen can be harvested for further analysis and culture downstream. Unlike other systems these are living cells and are not harmed in the manipulations.
This exciting technology is not inexpensive; Berkeley Lights has been selling these systems successfully into the biopharma segment, and it was clear that by exhibiting at AACR (they were also at ASHG in October 2018 in San Diego) they want to be known by the wider genomics community, for whatever it is they are working on next.
The most fun giveaway I’ve seen in some time
For today the prize goes to Canopy Biosciences, a new service offering for low-allele frequency mutation detection using UMI’s and their own analysis method out of Saint Louis MO. Not only did I attach the one below, the other one I obtained from their booth read “I wrote my thesis in Comic Sans”. Pay them a visit for these (and about six or seven more) at their booth #2650 and tell them Dale sent you.
Pingback: Single-cell interaction technology from Scribe Biosciences - Next Generation Technologist