Blog
May 30, 2023 / Oncology / Neuroscience / Immunology

Answering your questions about single cell analysis in clinical FFPE samples

Olivia Habern

You’re a clinical researcher studying rare diseases, and have been collecting patient samples from a small cohort over years. These samples are, out of necessity, formalin-fixed paraffin-embedded (FFPE)—and very precious.

You’re an oncologist studying the basis of therapeutic resistance in a clinical trial for solid tumors. You’ve collected pre- and post-treatment biopsies, preserving them via FFPE-processing, and now have correlative clinical outcomes from responders and non-responders to match to data from patient samples. But you need your experimental analysis of these samples to work the first time because, in a clinical trial, there are no “redos” for treatment or sample collection.

You’re also curious about the value that single cell analysis can bring to your studies—what questions it may help you answer about the biological heterogeneity of these samples, and the insights they contain regarding disease mechanisms and drug activity. But will single cell analysis in FFPE tissues, enabled by Chromium Single Cell Gene Expression Flex, really work in your samples of interest?

If this is you, we understand. You want to get the highest quality data and maximize insights from these valuable samples. You want a robust, reliable workflow that can ensure the best results, even from low-input and low-quality, aged samples. You want to honor the patients who have donated their samples to the purpose of advancing human health.

In a recent webinar, Q2 Solutions, a clinical research organization and 10x Genomics Certified Service Provider, addressed prevalent questions and concerns regarding single cell analysis in clinical FFPE tissues. They tested the Single Cell Gene Expression Flex assay on multiple clinically relevant FFPE tissues, including tumor samples from breast, lung, and ovaries, with varying quality scores and tissue input conditions. They also performed comparative analysis with bulk RNA-seq on matched FFPE samples.

We’ve condensed some of the insights from their study regarding sample requirements and key quality metrics into a Q&A below. Keep reading to learn how they confirmed that Flex can provide high-quality single cell gene expression data from FFPE tissues, then rewatch the webinar for yourself or explore data in their poster summarizing the study, featured at AACR 2023.

Sample requirements for FFPE single cell analysis 

How long can samples be stored and still provide high-quality single cell data?

FFPE is accepted within the field as a long-term tissue storage solution. It’s important to ensure that the tissue was properly embedded in formalin, and that it's been stored in appropriate long-term conditions as well.

If there is a concern regarding tissue quality, or if you want to establish some upstream QC metrics that would indicate how the sample might perform downstream, you can use the DV200 score, the percentage of RNA fragments greater than 200 nucleotides. This allows us to evaluate the quality of the sample prior to engaging in further downstream single cell processing.

What we’ve found is that DV200 status is a key predictor of downstream experimental success. In our data, we see a strong, positive, linear relationship between DV200 and our quality metric of choice, meaning higher DV200 equals better experimental outcomes. For the majority of metrics, including estimated number of cells, median genes per cell, and fractions of reads per cell, this finding is robust regardless of tissue type, though there may be some tissue-specific impact on median genes per cell. Specifically, our data demonstrates a DV200 score greater than or equal to 30 as a cut point—this is the point at which we extrapolate roughly 5,000 cells captured, and 60% fraction of reads per cell. This data combined with a strong historical precedent suggests a DV200 score of 30 is a reasonable threshold for tipping the scale towards high-quality data capture.

If DV200 score is the concern for your block, it may not be an overall rule out in terms of utilizing that sample for single cell analysis. One expected impact would be reduced cell capture efficiency for that sample. But we can engage in strategic pilots if there are samples of great interest, where optimizations may be implemented to allow for better downstream data acquisition.

*While we’ve found that DV200 is the best metric to assess FFPE sample quality for the Chromium Single Cell Gene Expression Flex workflow, if of interest, the tissue blocks used for this experiment were of varying ages: breast and lung blocks were created in 2018–2019, while ovary samples were from 1995.

How much FFPE tissue is required for input to the single cell workflow? 

For these blocks, cell yield variability is expected across tissue types and can even depend on the tissue’s position in the block. For the tissues we tested, we used multiple sections or curls from each block—there is an expectation that the composition of the tissue may be different for each curl. I'll also note that samples featuring lower quality DV200, specifically less than 30, are expected to show more variability. Finally, there is an expectation of cell loss throughout the cell manipulation process, from post-dissociation to post-hybridization.

With that in mind, we tested a range of sample inputs, from 1 x 25 μm curl, to 2 x 50 μm curls for a total of 100 μm of input. The tissues tested were breast, lung, ovary, and testes, and DV200 scores ranged from 3 to 66, though considered generally as either greater or less than 30. We found that for all input conditions, we were able to yield sufficient cells for downstream processing. Our preliminary analysis suggests that sample input as little as 1 x 25 μm curl of tissue with DV200 > 30 yields adequate cells for processing. Regarding cell counts, we’d recommend having at least 200,000 cells post-dissociation, and at least 60,000 cells post-hybridization. Lower inputs are also currently being evaluated.

What are your recommendations for the best way to dissociate FFPE tissue for single cell analysis? 

With FFPE tissue, there is a need to deparaffinize and rehydrate these samples prior to subsequent dissociation. There are multiple paths to dissociate FFPE tissues, for example, using a mortar and pestle or utilizing a more automated approach. In the interest of ensuring quality and assay throughput, we have implemented automated dissociation of FFPE tissue samples. The goal is to remove or replace any small scale, lower throughput processing steps, specifically those mortar and pestle dissociations, which are fairly low throughput. Manual methods can bring additional risks, such as operator variability or even over-lysing tissue samples. Automated approaches reduce and remove any of the variability or batch effects that can be quite critical for sample processing. For these experiments, we used the Miltenyi FFPE Tissue Dissociation Kit.

*10x Genomics has provided a demonstrated protocol for isolating cells from FFPE tissue sections here, with additional information regarding expected cell yield based on tissue type and number of curls.

Are there any tissues that may be more challenging for this workflow than the tissues you've examined already?

We evaluated a set of clinically relevant samples, including breast, lung, and ovary tumor tissues that featured specimens above and below recommendations for threshold FFPE DV200 score. This study therefore highlighted a particular group of tissues that may struggle from lower quality. But certainly there may also be other clinically relevant tissue samples that require additional optimizations before processing that sample type. We would recommend engaging in a pilot evaluation of those tissues to ensure no additional optimization may be required. And we have standard approaches for those pilot studies when that sample of interest may require additional assistance for dissociation, cell yield, or even accessibility of downstream data.

*10x Genomics has validated a number of different tissue types (including varying disease states) for FFPE single cell analysis, which can be reviewed here.

Can the single cell RNA-seq workflow with Gene Expression Flex be used on archived FFPE slides? 

There are a few considerations to note. Within the field, traditional curl sections on a slide are roughly 5 microns in thickness. While that may not be an immediate concern, typically, when sectioning below 25 microns, there may be a risk to whole cells being partially cut or even cut in half. This may lead to lower cell yield when isolating that section, and would impact cell capture efficiency. It may not allow for adequate cell yield for downstream processing. When cells are cleaved in half, or partially cut, along with cell yield, there may be concern for overall availability of sequenceable information. We want to be conscious of both of those risks.

We’d also want to ensure that the quality of the FFPE slide is intact, again, noting that there are recommendations in place to ensure proper mounting and storage of the slide, which is important for the best downstream processing. In order to enable sample processing of slides for FFPE single cell analysis, the slide needs to be scraped, so what we'd do is scrape away that material from the slide. Preliminary analysis performed at Q2 has shown that cells can be yielded from scraped slides, again, noting the variables that may impact the overall cell yield, such as quality and storage conditions.

Our recommendation, if there is interest in performing this processing from tissue on slides, is to allow for larger sections to be mapped onto the slides or to have access to multiple slides of 5 µm thickness, which would increase the cell yield.

Q2 is also engaging in strategic pilots to characterize data from lower input clinical FFPE samples. And we're certainly exploring slides and the overall stability conditions required for slides.

Expectations for data quality with FFPE single cell analysis

What is the expected size of RNA profiling libraries from the Gene Expression Flex assay?

There is an expectation for library size to be between 100 and 500 base pairs. And all libraries, regardless of quality, tissue type, or input amount, were found to be in that range, at an average of just below 300 base pairs.

*10x Genomics has provided a knowledge base article on this topic here for further reading, with additional recommendations for library concentration.

What is the recommended cell capture target and sequencing read depth to produce the most robust single cell data from FFPE tissue? 

We’d recommend targeting 10,000 cells to allow for a more full representation of cell types from the sample. For read depth, the minimum recommendation is 10,000 reads/cell, but we do recommend targeting 20,000 reads/cell for a more transcript level assessment.

Is there evidence for reproducibility across sites and users for the Gene Expression Flex assay? 

Yes. For the joint development project with 10x Genomics, we used a control testes sample provided by 10x to perform a matched comparison across sites. One sample was processed at 10x and three aliquots were provided to Q2 for replicate processing. For these testes samples, DV200 scores weren’t available, but the data does indicate a high-quality sample. In terms of cell capture targets, 10x targeted about 8,000 cells, whereas Q2 targeted about 10,000 cells for the sample. For read depth, 10x targeted about 6,000 reads per cell, while Q2 targeted higher at 20,000. We found that the data depicts high reproducibility across replicates at Q2 when compared across sites for key QC metrics, specifically looking at estimated number of cells, median genes per cell, total genes detected. These are some of the key QC metrics that we utilize to ensure successful processing.

Is single cell data from FFPE samples on par with or equivalent to bulk RNA-seq data from matched FFPE samples? 

We compared our single cell RNA-seq FFPE samples to matched bulk sequencing samples, to provide a comparative assay. And I would say these are biologically consistent readouts. What we’ve been able to show is that, even with these challenging input types, FFPE in particular, we are capturing similar populations of cells in linearly related quantities across the two platforms. They're wildly different in terms of the number of cells they can assay and the ability to assess rare cell populations. But what's really nice is the consistency in cell capture across the two systems and their relatedness in estimating the various cell types in the sample. You can start doing some cool development on your single cell workflows and maybe transition into other workflows as needed for various reasons.

Watch the webinar from Q2 Solutions to find even more details about the results of their studies on FFPE tissues using Chromium Single Cell Gene Expression Flex.

Flex makes single cell analysis possible where it wasn’t before, giving clinical researchers a reliable, robust workflow to uncover critical cellular insights from precious samples, whether biobanked FFPE clinical samples or even fresh or PFA-fixed tissues. Keep exploring what Flex can do here →

About the author:

Olivia received a BA in Molecular & Cell Biology with an emphasis in Immunology and an English minor from UC Berkeley. She began writing as an undergraduate for the High Performance Computing program, interviewing Berkeley professors, postdocs, and graduate researchers about how they were using computing to analyze their data and model experimental systems, from weather forecasting to nuclear reactions. This inspired her interest to tell stories about innovative science and the technology that makes it possible, which led her to 10x Genomics where she has been writing for the last 5 years.
Olivia received a BA in Molecular & Cell Biology with an emphasis in Immunology and an English minor from UC Berkeley. She began writing as an undergraduate for the High Performance Computing program, interviewing Berkeley professors, postdocs, and graduate researchers about how they were using computing to analyze their data and model experimental systems, from weather forecasting to nuclear reactions. This inspired her interest to tell stories about innovative science and the technology that makes it possible, which led her to 10x Genomics where she has been writing for the last 5 years.