Question: What QC is available before, during, and after running the Flex Gene Expression assay with FFPE samples?
Answer: There are several opportunities to QC samples before, during, and after performing the Flex Gene Expression assay with FFPE samples. Listed below are QC steps- each of which are expanded upon in the body of the article.
- DV200, as determined by bulk RNA extraction(s) from FFPE curls
- Count and QC nuclei suspension(s) post-dissociation (recommended range is 400,000-2x10^6 nuclei per Hybridization)
- Post-Hybridization QC prior to Chip loading
- Hero metrics and plots in Web Summary file (Cell Ranger output)
Before Running the Assay:
DV200 Assessment
We found that FFPE blocks that performed well with Visium CytAssist Spatial Gene Expression for FFPE experiments were highly likely to perform well with the Fixed RNA Profiling assay too. Upstream of Visium experiments with FFPE samples, we recommend performing a bulk RNA extraction to determine the DV200 (the percentage of RNA fragments >200 nucleotides) of your block(s) of interest. For Visium CytAssist Spatial Gene Expression for FFPE experiments, we determined that tissue samples with a DV200 of >30% were more likely to yield robust and reproducible results. Therefore, prior to starting the Sample Preparation from FFPE Tissue Sections for GEM-X Flex Gene Expression (i.e., FFPE Demonstrated Protocol), users can perform DV200 assessment to screen FFPE blocks. This will enable users to determine which block (of many) may be more ideal to run with the Flex Gene Expression assay.
- For details on how to perform bulk RNA extraction, third-party items needed, and how to interpret RNA traces, please refer to the Visium CytAssist Spatial Gene and Protein Expression for FFPE Tissue Preparation Guide Demonstrated Protocol (CG000660)
- For matched Gene Expression Flex and Visium data, see this Datasets page.
It is important to note that DV200 is not a perfect predictive tool as RNA quality is not the only factor that influences performance of FFPE experiments. For instance, sample complexity and nuclei yields also play a role in the ability to generate good-quality data. For additional details on DV200, please see the Appendix of the Tissue Preparation Guide linked above.
Post-Dissociation QC
After tissue dissociation, count and QC nuclei suspensions using fluorescent dye such as PI and an automated cell counter. Be mindful of automated cell counter specifications, such as optimal cell concentrations and size limitations. When counting, count bright particles as cells/nuclei and exclude dimly fluorescent particles (Figure 1). For additional details on cell counting best practices, see this Knowledge Base Article.
Figure 1. Fluorescent image of nuclei after tissue dissociation where nuclei considered for counting and nuclei not considered for counting are noted. Bright particles are gated for counting while dim particles are not.
While reviewing images, suspensions should appear uniform and with minimal aggregates (Figure 2A). Certain levels of debris are expected, which are typically minimized after the Post-Hybridization Washes and filtering. Images (brightfield and fluorescent) of a representative nuclei suspension after Dissociation are shown below. This can be found in the FFPE Demonstrated Protocol (linked above).
During the Assay:
Post-Hybridization Washes QC
After completing Post-Hybridization Washes and preparing to load the 10x Chip, samples will be counted, and QC’d. Use PI and an automated cell counter to QC and count nuclei.
Shown below is a representative sample after Dissociation and Post-Hybridization Washes (Figure 2). As mentioned above, suspensions appear more uniform with fewer clumps and debris after the Post-Hybridization Washes and subsequent filtering using the validated 30 µm filters listed in the Flex Gene Expression User Guides (Figure 2B; linked here). Having a uniform suspension without debris or clumps is critical for minimizing the likelihood of experiencing a clog or wetting failure during GEM Generation.
Figure 2. Cellaca brightfield (top) and fluorescent (bottom) images of nuclei A. after dissociation and filtering (left) and B. after Post-Hybridization Washes (right). Suspensions are cleaner after Post-Hybridization Washes.
Additional images are included in Flex Public Datasets under “Input Files” (Public Datasets can be found here).
After Running the Assay:
Web Summary File Metrics
Below are observations made from reviewing FFPE Web Summary files and steps to use these metrics to QC the data.
Figure 3. Examples of good (green outline; DV200=67%), acceptable (yellow outline; DV200=40%), and poor (red outline; DV200=30%) Gene Expression Flex data as demonstrated by the Barcode Rank Plot (top), Cell recovery (text), and t-SNE Plot (bottom) generated from human Reactive Lymph Node (LN), Healthy Breast, and Healthy Heart FFPE samples.
Barcode Rank Plot
For the FFPE assay, we define a good-quality Barcode Rank Plot as having a sharp inflection point and a relatively low number of UMIs in non-cell barcodes. An example of this is shown in the human reactive lymph node (green outline; Figure 3) Barcode Rank Plot. This is in comparison to the human healthy breast (yellow) and human healthy heart (red) plots that have a curve, rather than steep inflection point.
Some FFPE samples won’t have a distinct inflection in their Barcode Rank Plots, as this plot is influenced by cell complexity (e.g., “Median UMIs per Cell”) and yield (i.e., “Estimated Number of Cells”). FFPE is a particularly challenging sample type for tissue dissociation and cell recovery. Therefore, cell yield and cell recovery are often lower or more variable than freshly isolated and fixed samples. For this reason, we recommend targeting the maximum number of cells supported by the workflow to maximize likelihood of success.
“Estimated Number of Cells”
We have found cell recovery with FFPE samples may be variable or lower than with other sample inputs. This may be due to low complexity and the cell calling algorithm. Because of this, we aim for >50% recovery based on “Estimated Number of Cells” as a fraction of Targeted Cells. For instance, we see that the lymph node sample (green outline) recovered ~80% of targeted cells while breast and heart samples recovered much fewer (<50%, yellow and red outlines, respectively).
t-SNE Projection of Cells Colored by Cluster
Typically, we expect the t-SNE Projections to generate distinct and separate clustering like what’s illustrated in the lymph node sample (green outline). It’s important to note that the number of clusters, distance separating clusters, and the UMI counts within each cluster may vary in heterogeneous samples. In line with this, you may observe a subset of clusters with much higher UMI counts while other clusters may be of lower complexity. In samples where there’s little diversity in cell types or there’s high background, clustering may be poorer. For example, while the breast t-SNE Plot has acceptable clustering (yellow outline), the heart sample has poor separation of clusters (red outline). The distance between clusters and the number of assigned clusters give an idea of how much noise is in the data and diversity of cell types of the sample.
Products: Flex Gene Expression