Question: Are healthy bone FFPE tissues compatible with Visium HD Spatial Gene Expression?
Answer: Bone is a challenging tissue with the Visium HD Spatial Gene Expression Assay. Acceptable data quality is dependent on many factors, including species (mouse vs. human), bone type, sample quality, post-mortem interval (PMI), and tissue processing. These challenges stem from the difficulty in obtaining high quality RNA, the necessity of decalcification, and high rates of tissue detachment. Based upon limited testing, we found that mouse bone showed expected spatiality, but human bone had relatively poor complexity and sensitivity even when implementing these protocol modifications. We also observed lower DV200 values (~30 - 35%) for both the mouse and human bone samples tested here, further reflecting the negative impact of processing steps like decalcification on sample quality.
The FFPE mouse and human bone samples tested on Visium HD were fixed, decalcified, processed, and sectioned following the guidance provided on pages 1-6 in the Analysis of Bone Tissue using Xenium v1 In Situ Gene Expression Assay Technical Note (CG000816).
Figure 1. Methods overview for processing of mouse and human bone.
Tissue processing methods described in CG000816 may require further optimization for which 10x Genomics cannot provide additional guidance, guarantee assay performance, or share additional details from internal experiments. Since data quality from bone samples may be variable, it is listed as a ”challenging tissue” type in our FFPE Tested Tissues List. We have minimal recommendations for working with human bones based on experimental data generated from tibia & femur. Bone samples outside of those noted below have not been tested. Before proceeding with a Visium HD experiment with bone samples, we would recommend assessing the DV200 and carrying out optional morphology QC (DAPI and/or H&E staining) following the steps in the Visium HD FFPE Tissue Preparation Handbook (CG000684), and performing a small pilot experiment with bone tissue before scaling up.
Mouse Bone
Table 1 and Figures 2-4 show data from representative web summary files when processing mouse femur (DV200 - 32%) and mouse tibia (DV200 - 35%) on Schott Nexterion Slide H - 3D Hydrogel Coated Slides following guidance provided in the Xenium Technical Note (CG000816), Visium HD FFPE Tissue Preparation Handbook (CG000684), and Visium HD User Guide (CG000685).
- The HD Spatial Gene Expression libraries generated for these mouse bone samples were of good quality, as Valid Barcodes and Valid UMIs were high. Sequencing performed as expected with high Q30 scores.
- Probe set mapping metrics (Reads Mapped to Probe Set, Reads Mapped Confidently to Probe Set, and Reads Mapped Confidently to the Filtered Probe Set) were high, indicating that probe hybridization and ligation performed as expected.
- Fraction Reads in Squares Under Tissue was ~93%. Genomic DNA was < 2%. Both the mouse bone samples had good assay sensitivity. Mouse femur had 1844 mean reads per 8 μm bin and 223 UMIs per 8 μm bin. Mouse tibia had 2093 mean reads per 8 μm bin and 289 UMIs per 8 μm bin.
- UMI maps show that the regions of bone with high UMIs (>400 UMIs) primarily localize to the bone marrow. We observe low UMI counts in the bone and muscle. Therefore, the majority of UMIs originate from the bone marrow of the sample with minimal signal from the bone itself.
- Graph-based spatial clustering, UMAP projections, and the plots for marker genes represent the heterogeneous composition of cell types in the mouse bone samples.
Table 1. Summary of data metrics for mouse tibia and femur.
| Data Metric | Mouse Femur | Mouse Tibia |
| Valid Barcodes | 92.6% | 93.6% |
| Valid UMIs | 99.8% | 99.8% |
| Q30 Bases in Barcode | 96.3% | 96.4% |
| Q30 Bases in Probe Read | 96% | 95.9% |
| Q30 Bases in UMI | 96.7% | 96.7% |
| Sequencing Saturation | 85.8% | 84.3% |
| Reads Mapped to Probe Set | 99.2% | 99.4% |
| Reads Mapped Confidently to the Filtered Probe Set | 98.7% | 98.8% |
| Fraction Reads in Squares Under Tissue | 92.8% | 94.5% |
| Fraction UMIs from Genomic DNA | 1.3% | 2.1% |
| Mean Reads per 8 µm Bin | 1844.6 | 2093.8 |
| Mean UMIs per 8 µm Bin | 223.1 | 289.9 |
FFPE Mouse Femur
Figure 2. Key web summary file plots generated from mouse femur tissue. (A) H&E image (B) UMI map (C) Graph-based clustering (D) UMAP
FFPE Mouse Tibia
Figure 3. Key web summary file plots generated from mouse tibia tissue. (A) H&E image (B) UMI map (C) Graph-based clustering (D) UMAP
Figure 4. Plots showing canonical markers overlaid on mouse tibia tissue. (A) Igkc is expressed in plasma cells. (B) Col1a1 is expressed in cartilage and bone (C) Bpgm is expressed in red blood cells in bone marrow. (D) Actn3 is expressed in muscle.
Human Bone
Table 2 and Figures 5 and 6 show data from representative web summary files when processing human femur (DV200 - 30%) and tibia (DV200 - 31%) on Schott Nexterion Slide H - 3D Hydrogel Coated Slides following guidance provided in the Xenium Technical Note (CG000816), Visium HD FFPE Tissue Preparation Handbook (CG000684), and Visium HD User Guide (CG000685).
- Human bone samples are particularly challenging due to the longer decalcification process which typically takes 3-4 weeks. The longer decalcification required for human bone processing can impact RNA quality and result in varying levels of gDNA.
- The HD Spatial Gene Expression libraries generated for these human bone samples were of good quality, as Valid Barcodes and Valid UMIs were high. Sequencing performed as expected with high Q30 scores.
- The Reads Mapped to Probe Set metric is high (~97%). The Reads Mapped Confidently to Probe Set (>88%) and Reads Mapped Confidently to the Filtered Probe Set (~83%) metrics are a bit lower than in mouse bone samples but still within the acceptable range.
- Fraction Reads in Squares Under Tissue was ~88%. While the human femur had ~9% of estimated UMIs from genomic DNA (gDNA), the human tibia had ~60% of estimated UMIs from gDNA, informing us that a substantial portion of reads in the human tibia may be originating from off-target hybridization to gDNA. Samples with poorer quality often tend to also have higher levels of genomic DNA as seen in these samples .
- Human bone samples had poorer assay sensitivity compared to mouse bone. Human femur had 533 mean reads per 8 μm bin and 14 UMIs per 8 μm bin. Human tibia had 592 mean reads per 8μm bin and 2.9 UMIs per 8 μm bin. Lower sensitivity in these samples also severely impacted the resolution of clusters in the UMAP (data not shown).
- The UMI maps show that the regions of the bone with relatively high UMIs (>30 UMIs) primarily localize to the bone marrow. We observe very low UMI counts in the bone and muscle. Therefore, the majority of UMIs originate from the bone marrow of the sample with minimal signal from the bone itself.
Table 2. Summary of data metrics for human tibia and femur.
| Data Metric | Human Femur | Human Tibia |
| Valid Barcodes | 92.9% | 89.9% |
| Valid UMIs | 99.8% | 99.8% |
| Q30 Bases in Barcode | 97.9% | 97.9% |
| Q30 Bases in Probe Read | 97.8% | 97.9% |
| Q30 Bases in UMI | 97.3% | 97.2% |
| Sequencing Saturation | 97.8% | 99.3% |
| Reads Mapped to Probe Set | 97.6% | 97.8% |
| Reads Mapped Confidently to the Filtered Probe Set | 90.3% | 84.8% |
| Fraction Reads in Squares Under Tissue | 88.6% | 88.2% |
| Fraction UMIs from Genomic DNA | 9.4% | 60.8% |
| Mean Reads per 8 µm Bin | 533.3 | 592 |
| Mean UMIs per 8 µm Bin | 14.1 | 2.9 |
FFPE Human Femur
Figure 5. Key web summary file plots generated from human femur tissue. (A) H&E image (B) UMI map (C) Clustering
FFPE Human Tibia
Figure 6. Key web summary file plots generated from human tibia tissue. (A) H&E image (B) UMI map (C) Clustering
In this article, we reviewed our findings and challenges when working with mouse and human bone samples. Based on limited internal data from healthy FFPE mouse femur and tibia tissue, we found that mouse bone can generate good Visium HD Spatial Gene Expression data. This is shown in Figure 4 where canonical markers of bone marrow (4A, 4C), cartilage and bone (4B) and muscle (4D) can be identified in the expected regions of the tissue. Thus, it may be possible to generate quality Visium HD data from mouse bone samples by following the sample preparation guidance provided in the Xenium Technical Note upstream of the Visium HD FFPE Tissue Preparation Handbook & Visium HD User Guide, and utilizing Nexterion Hydrogel tissue slides. Please note that additional optimization on targeted regions of mouse bone samples may be required, depending on experimental goals. In contrast, we determined that it was more challenging to generate quality gene expression data with human bone samples. Even with modifications provided in the Xenium Technical Note, spatiality and sensitivity was poor. As data quality may be variable with bone samples, we recommend performing sample QC steps and an initial pilot experiment before committing to larger experiments.
Product: Visium HD