Here we present protocols that enable isolation of stromal cells from murine bone, bone marrow, thymus and human thymic tissue compatible with single-cell multiomics.
Single-cell sequencing has enabled the mapping of heterogeneous cell populations in the stroma of hematopoietic organs. These methodologies provide a lens through which to study previously unresolved heterogeneity at steady state, as well as changes in cell type representation induced by extrinsic stresses or during aging. Here, we present step-wise protocols for the isolation of high-quality stromal cell populations from murine and human thymus, as well as murine bone and bone marrow. Cells isolated through these protocols are suitable for generating high-quality single-cell multiomics datasets. The impacts of sample digestion, hematopoietic lineage depletion, FACS analysis/sorting, and how these factors influence compatibility with single-cell sequencing are discussed here. With examples of FACS profiles indicating successful and inefficient dissociation and downstream stromal cell yields in post-sequencing analysis, recognizable pointers for users are provided. Considering the specific requirements of stromal cells is crucial for acquiring high-quality and reproducible results that can advance knowledge in the field.
In the healthy adult, de novo production of blood cells occurs in the bone marrow and the thymus. Stromal cells at these sites are essential for maintenance of hematopoiesis, but stroma constitutes less than 1% of the tissue1,2,3,4. Obtaining pure isolates of hematopoiesis supporting stroma therefore constitutes a significant challenge, particularly for single-cell multiomics that requires expedient processing to obtain samples of high quality. Components of different digestion cocktails may interfere with certain steps in multiomics analysis5,6. The protocols presented here detail the isolation of a wide variety of stromal cells from bone marrow and thymic tissues.
Perturbations of stromal constituents in both bone marrow and thymus result in profound disruption in blood cell development and can result in malignancies7,8,9. Hematopoiesis supporting stroma is damaged following cytotoxic conditioning and bone marrow transplantation, resulting in reduced secretion of cytokines and growth factors that sustain hematopoietic stem and progenitor cells (HSPCs)2,10,11. Furthermore, aging affects bone marrow and thymus stromal cells likely contributing to aged hematopoietic phenotypes. The thymus is the first organ to undergo extensive age-associated involution. Fat and fibrotic tissue start replacing T cell supportive stroma as early as the onset of puberty12,13. In the bone marrow, adipocyte content increases with age and the vascular and endosteal niches are significantly remodeled14,15,16.
To enable study of hematopoiesis supportive stroma across multiple stress states and in the case of the thymus of both human and murine tissue, we have optimized previously published digestion protocols1,2,8,17,18. These protocols ensure efficient and reproducible isolation of cells, and they are compatible with single-cell RNAsequencing (scRNAseq) and other types of multiomics.
All work with human tissue was conducted after approval by the Massachusetts General Hospital Internal Review Board (IRB). All animal procedures were conducted in accordance with the Massachusetts General Hospital Institutional Animal Care and Use Committee (IACUC) guidelines. C57Bl/6 mice, 8-10 weeks old, and both males and females, were used for the present study. The animals were obtained from a commercial source (see Table of Materials).
1. Preparation of murine thymic tissue
2. Preparation of human thymic tissue
3. Preparation of murine bone and bone marrow tissue
NOTE: Bone and bone marrow fractions are prepared in two separate digestion reactions to obtain maximum purity of stromal cells and optimal dissociation of tissues. The samples can be pooled after the digestion steps to be sorted as one stromal compartment.
These protocols yield reproducible stromal cell varieties from the thymus and bone marrow suitable for flow cytometric analysis, as well as single-cell multiomics, such as scRNA sequencing. Murine thymic tissue undergoes significant remodeling in response to stressors, such as the cytotoxic conditioning that precedes bone marrow transplantation or the natural aging process. As a consequence, thymic cellularity is drastically reduced in both of these settings (Figure 1A). While a thymus from an 8-week-old wild-type mouse contains approximately 100 million cells, the cellularity of a 2-year-old mouse can be expected to be half of that, and a thymus 4 days post-irradiation and bone marrow transplantation can be as low as 5,00,000 cells (Figure 1A).
Cytotoxic conditioning will preferentially ablate hematopoietic cells, so the stromal compartment will be proportionally enriched (Figure 1B). However, the overall low cellularity will make isolating sufficient numbers of thymic stromal cells challenging. It may, therefore, be necessary to pool animals to obtain cell numbers compatible with downstream multiomics applications and to minimize the loss of material wherever possible. To perform multiomics on thymic stromal cells, fluorescence-activated cell sorting (FACS) of live, single, CD45-Ter119- cells (Figure 1C) will ensure that sufficient numbers of cells can be sorted from most stress conditions.
To preserve the heterogeneity of the stroma compartment, we typically do not sort on a single positive stroma marker but gate out hematopoietic cells. Including positive stromal cell markers is still recommended for analysis to ensure that proper tissue dissociation has been achieved. However, FACS isolation is imperfect, and as determined by scRNAseq, 30%-50% of the sorted cells will be of hematopoietic origin (Figure 1D,E). This can be improved upon by magnetically depleting CD45+ cells prior to FACS; however, this is not recommended for conditions of reduced thymic cellularity (Figure 1A) as the loss of material that comes with the extra processing involved in magnetic depletion results in stroma yields too low for downstream applications.
For the isolation of human thymic stroma, the most critical step is the enzymatic digestion. If the dissociation step is rushed, the overall stromal cell yield, as well as stromal cell subtype representation, will be severely diminished (Figure 2A). It is also recommended to always include the markers of thymic endothelium and epithelium to ensure that tissue digestion has been optimal, as simply gating negative for hematopoietic markers will not be sufficient to gauge actual stromal cell release. This is exemplified in Figure 2A, where the percentage of cells negative for hematopoietic markers is higher in the sample digested for 30 min compared to the sample incubated for 90 min. However, as seen in the consecutive gate, the yield of endothelial and epithelial cells is very poor in the shorter digestion protocol (Figure 2A).
For FACS isolation of human thymic stromal cells for multiomics applications, staining with a hematopoietic lineage antibody cocktail, as well as CD4 and CD8 antibodies is suggested, in addition to the pan-hematopoietic marker CD45 and the erythroid marker CD235a (Figure 2B) to improve stromal cell enrichment. Thus, for human multiomics analysis of thymic stroma, live, single, CD45-CD235a-Lineage-CD4-CD8- cells should be FACS sorted (Figure 2B). As seen for scRNAseq of murine samples, this still results in contamination of 30%-50% hematopoietic cells, but this still gives acceptable resolution of the human thymus stromal compartment (Figure 2C,D).
For bone and BM stroma, the same strategy was used as for thymic tissue; stromal cells are isolated by gating on cells negative for hematopoietic markers, and further stromal cell markers are included as digestion controls (Figure 3). The dissociation of calcified tissue like bone results in a significant accumulation of debris. The addition of calcein as a viability dye will allow the exclusion of most of this debris, which otherwise falsely inflates sorted numbers and decreases gated population frequencies (Figure 3). This is exemplified in the FACS plots comparing gating on all cells and just gating on Calcein+ (Figure 3). As seen for human thymic stromal cells, the percentage of hematopoietic marker-negative cells is not a reliable readout for bone and BM tissue dissociation efficiency (Figure 3). While the CD45-Ter119- gate can contain an abundance of cells in a sample where the digestion step failed, further analysis of typical bone and BM stromal cell markers CD31, CD140a, and CD51 clearly demonstrate poor stroma release (Figure 3 and Figure 4).
For FACS isolation of murine bone and BM stromal cells, using the gating strategy is suggested as in Figure 3, where we only sort CD45-Ter119-Calcein+ but monitor the forward and side scatter properties and stromal cell marker expression of the sorted population to ensure optimal digestion (Figure 3A). The combination of hematopoietic lineage-positive magnetic cell depletion of the bone marrow prior to FACS and sorting on the shown gates results in single-cell sequencing of cells that are >90% stroma. An example of such a scRNA seq run (Figure 5A,B) illustrates this low hematopoietic contamination after transcriptome-based annotation of the analyzed cells (examples of hematopoietic marker genes Figure 5C). This is in stark contrast to the thymic stromal preparations where magnetic cell depletion was not performed before the sorting step due to the limited starting material in post-transplant thymic tissue (Figure 1F).
Figure 1: Flow cytometric and single-cell RNA sequencing analysis of murine thymic stroma. (A) Quantification of thymocyte numbers in 8-week-old mice (Control), 8-week-old mice lethally irradiated and transplanted 4 days prior (Transplant) and 2-year-old mice (Aged). (B) Representative flow cytometry plots demonstrating the enrichment in stromal cell subtypes 4 days following lethal irradiation and transplantation. (C) Flow cytometry gating strategy to sort murine thymic stroma for single-cell multiomics analysis. (D) UMAP of murine thymus indicating stroma (blue) and hematopoietic cells (red). (E) UMAPs showing murine hematopoietic marker genes in thymus samples. (F) Quantification of murine stromal and hematopoietic cells in thymus scRNAseq samples. Please click here to view a larger version of this figure.
Figure 2: Flow cytometric and single-cell RNA sequencing analysis of human thymic stroma. (A) FACS plots showing representative results of the impact of different digestion times on the release of human thymic stromal cell populations. (B) Flow cytometry gating strategy to sort human thymic stroma for single-cell multiomics analysis. (C) UMAP of human thymus indicating stroma (blue) and hematopoietic cells (red). (D) UMAPs showing human hematopoietic marker genes in thymus samples. (E) Quantification of human stromal and hematopoietic cells in thymus scRNAseq samples. Please click here to view a larger version of this figure.
Figure 3: Murine bone marrow stromal markers and Calcein are helpful to estimate dissociation efficiency. (A) Representative flow cytometry analysis of CD45–Ter119– murine bone marrow stromal cells gated on all cells and Calcein+ cells. With subsequent CD31– gating showing a higher proportion of CD140a+CD51+ cells when selecting Calcein+ cells. (B) Flow cytometry analysis representative of a failed bone digestion with missing CD140a+CD51+ cells. Please click here to view a larger version of this figure.
Figure 4: Bone marrow stroma gating strategy. (A) Flow cytometry gating strategy to sort murine bone marrow stroma for single-cell multiomics. (B) Flow cytometry analysis representative of a failed bone marrow digestion with missing CD140a+CD51+ cells. Please click here to view a larger version of this figure.
Figure 5: Bone marrow single-cell RNA sequencing with stromal and hematopoietic cell composition. (A) UMAP of murine bone and bone marrow indicating stroma (blue) and hematopoietic cells (red). (B) Quantification of stromal and hematopoietic cells in murine bone and bone marrow samples. (C) UMAP of hematopoietic marker genes in murine bone marrow samples. Please click here to view a larger version of this figure.
Stromal cells in hematopoietic organs are critical for normal blood production and hematopoietic stroma perturbations can result in severe impairments in hematopoietic maintenance and response to stress9,23,24. Insight into hematopoietic stromal cells is essential for understanding hematological diseases7,9,10,24 and for the ability to develop therapeutics to combat them25,26. Reproducibly separating hematopoietic stroma from parenchyma is however a challenge, particularly with the advent of single-cell multiomics which require high-quality sample preparations while maximizing stromal cell yield and variety. The protocols described here have been optimized to be compatible with most downstream multiomics applications while meeting the challenges of isolating hematopoietic stromal cells1,2,19.
The first issue to consider for isolation of hematopoietic stromal cells is dissociation conditions. Optimal dissociation is critical to release stroma cells from the tissue to reflect full heterogeneity of the stroma compartment1,2. Choosing the right dissociation mix depends on the experimental scope and downstream protocols. We would recommend preparing dissociation mixes as freshly as possible and keeping them on ice until use. Some components are hard to dissolve such as dispase and might require more time for pipet mixing. Vortexing is not recommended as this may disrupt the native conformation of the digestion enzymes, ultimately leading to their inactivation. Importantly, EDTA will inhibit enzymatic activity and should be avoided during active digestion steps, but can be employed to stop the reaction if the enzyme containing buffer cannot be removed promptly. In addition to the enzyme mixture itself, two points are critical to efficient dissociation of tissue, (1) thorough crushing and cutting to increase surface area and, (2) sufficient incubation time with the digestion enzymes. Human thymic tissue is for instance significantly harder to dissociate than murine tissue and digestion times must therefore be longer. If adjustment is not made for human tissue, stromal yield and cell variety will be decreased (Figure 2A). Dissociation enzymes require an optimal temperature of 37 °C which inevitably influences gene expression27. Although cold or lower temperature dissociation would be desirable to conserve cell states, currently there are no commercially available cold-active dissociation enzyme combinations suitable for bone and bone marrow. For any tissue that requires digestion, the heterogeneity and the percent contribution acquired from different cell types will invariably change with the digestion time, enzyme cocktail composition and processing. Downstream of these initial steps, heterogeneity and representation will be further limited by filtration steps such as nozzle sizes during sorting and capture of challenging cell-types (e.g., adipocytes from stroma) and fluidic device channel diameters. Thus, when comparing datasets, we should be mindful that these differences in protocols can impact the representation of subpopulations. It logically follows that we must be vigilant about keeping processing steps equal when measuring the impact of perturbations in stromal systems.
While the cost of sequencing has been and is decreasing, single cell technologies are often limited by the number of cells that can be analyzed. To optimize the capacity of each reagent kit, hematopoietic cell magnetic depletion was employed whenever possible. This enables the avoidance of a high number of contaminant cells which are much more numerous than stroma in both bone marrow and thymic sample fractions. However, if the starting material is limited, as in the case of post-transplant thymic tissue, the increased, non-specific loss of cells that magnetic depletion strategies invariably incur, may make this inadvisable. In addition, during preparation and digestion of the tissues, crushing, cutting and enzymatic release of extracellular matrix components generates a large amount of debris. This type of contaminant, if left in the sample for analysis, can cause clogging issues in downstream processing, cause negative wells for single cell index sorting and highly inflate parent gate percentages in analysis. Calcein staining was used to gate positive for cells/live cells. Calcein comes in different colors which makes it highly adaptable to your staining matrix. When working with calcein be careful to not subject the dye to repeated freeze-thaw cycles (i.e., always prepare fresh and keep at room temperature after dissolving).
The ability to multiplex samples based on antibodies barcoded with DNA oligos28 allows increased utility of single cell sequencing reactions on platforms such as 10x Genomics. The hashtagging antibodies designed for sample multiplexing come with different targets for cell surface molecules or the nuclear membrane and specificity for both mouse and human. While the cell hashtagging system has been used successfully for multiple tissues, experiments combining digestion enzyme mixes with subsequent use of hashtags can lead to drastically reduced recovery. We have attempted to include additional washing steps post-dissociation and added EDTA to inhibit the activity of the digestion enzymes. However, even with these steps, expect lower hashtag identification than hematopoietic cells. It was found that with enzymatic digestion protocols, the hashtag capture rates were between 30%-70%. This occurs despite the fact that the same antibody clone with a fluorochrome conjugate labeled >90% of stromal cells in the same population. Any experimental design pairing enzymatic digestion of tissue with hashtagging should be piloted and caution taken to ensure enzymatic inhibition.
The authors have nothing to disclose.
We were supported with expert technical assistance by the HSCI-CRM Flow Cytometry facility at Massachusetts General Hospital and the Bauer Core Facility at Harvard University. T.K and K. G were supported by the Swedish Research Council and C.M. by the German Research Foundation. We thank Sergey Isaev and I-Hsiu Lee for assistance in analysis of single-cell RNA sequencing data.
0.25% Trypsin-EDTA | Thermo Fisher Scientific | 25200-072 | |
7AAD (7-aminoactinomycin D) | BD Biosciences | 559925 | |
Anti-Human Lineage Cocktail 3-FITC | BD Biosciences | 643510 | |
Bovine Serum Albumin | Millipore Sigma | A9647 | |
C57Bl/6 mice | Jackson | 664 | Males or females, 8-12 weeks old |
Calcein | Fisher Scientific | 65-0853-78 | |
Collagenase IV | Millipore Sigma | C5138 | |
Corning Sterile Cell Strainers, White, Mesh Size: 70 µm | Fisher Scientific | 08-771-2 | |
DAPI (4',6-Diamidino-2-Phenylindole, Dilactate) | Biolegend | 422801 | |
Dispase II | Thermo Fisher Scientific | 17105041 | |
Dnase I Solution | Thermo Fisher Scientific | 90083 | 2500 U/mL |
Easysep mouse streptavidin RapidSpheres Isolation kit | StemCell Technologies | 19860 | |
Fetal Bovine Serum | Gibco | A31605-01 | Qualified One Shot |
Human Fc Block | BD Biosciences | 564220 | |
Liberase TM | Millipore Sigma | 5401127001 | Research Grade |
Medium 199 | Gibco | 12350 | |
Mouse anti-human CD235a-BV77 | BD Biosciences | 740785 | |
Mouse anti-human CD31-PE/Dazzle594 | Biolegend | 303130 | |
Mouse anti-human CD45-BV77 | Biolegend | 304050 | |
Mouse anti-human CD4-BV605 | BD Biosciences | 562658 | |
Mouse anti-human CD66b-FITC | BD Biosciences | 555724 | |
Mouse anti-human CD8-APC/Cy7 | BD Biosciences | 557760 | |
Mouse anti-human EpCam-BV421 | Biolegend | 324220 | |
Protector RNase Inhibitor | Millipore Sigma | 3335402001 | |
Rat anti-mouse CD105-PE /dazzle594 | Biolegend | 120424 | |
Rat anti-mouse CD11b-Biotin | Biolegend | 101204 | |
Rat anti-mouse CD140a-APC | Fisher Scientific | 17-1401-81 | |
Rat Anti-Mouse CD16/CD32 (Mouse BD Fc Block) | BD Biosciences | 553142 | |
Rat anti-mouse CD31-BUV737 | BD Biosciences | 612802 | |
Rat anti-mouse CD31-BV421 | Biolegend | 102424 | |
Rat anti-mouse CD3-Biotin | Biolegend | 100244 | |
Rat anti-mouse CD45.2-Biotin | Biolegend | 109804 | |
Rat anti-mouse CD45-PE/Cy7 | Biolegend | 103114 | |
Rat anti-mouse CD45-PE/Cy7 | Biolegend | 103114 | |
Rat anti-mouse CD45R/B220-Biotin | Biolegend | 103204 | |
Rat anti-mouse CD51-PE | Biolegend | 104106 | |
Rat anti-mouse EpCam-BV711 | BD Biosciences | 563134 | |
Rat anti-mouse Ly-6A/E(Sca-1)-AF700 | Biolegend | 108142 | |
Rat anti-mouse Ly-6G/Ly-6C(Gr1)-Biotin | Biolegend | 108404 | |
Rat anti-mouse Ter119-Biotin | Biolegend | 116204 | |
Rat anti-mouse Ter119-PE | Biolegend | 116208 | |
Rat anti-mouse Ter119-PE/Cy7 | Biolegend | 116222 | |
Stemxyme | Worthington Biochemical | LS004107 |