Donor-dependent variation of human umbilical cord-derived MSC in proliferation and immune modulation

Cultured human umbilical-cord-derived MSC (huMSC) from 32 donors at passage 5 (P5) were characterized by their ability to differentiate into bone, cartilage, and fat cells (Fig. 1A), as well as carrying cell surface markers CD 44, 73, 90, 105, but not CD19, 34, 45, 11b, or HLA-DR (Fig. 1A), which were basic common features for MSCs. However, these 32 lines of P5 MSCs demonstrated variability in their proliferation rates (Fig. 1B and Supplementary Table S2). To investigate whether immunomodulation function of these 32 different MSC lines were also variable, we first used an in vitro culture model, where a mouse microglial cell line BV2 cells were cultured for 48 h in conditioned medium (CM) collected from 32 different lines of P5 huMSCs. The 32 lines of huMSCs were inoculated at the same density and cultured for 72 h for collection of CM. The suppressive effect of CM from each line of huMSCs on BV2 was deemed as an index for capability of microglia/immune suppression (Fig. 1B). Results showed that the suppressive indices (SI) of MSCs from different donors were variable, ranging from 0.256 to 0.721 (Fig. 1B). We further determined correlations between huMSC proliferation rates and their BV2 immune suppressive indices as well as potential influences from gender to immune modulation function of huMSCs, and found that neither proliferation rate nor gender appeared to significantly influence the ability of huMSCs to suppress BV2 cells (Fig. 1C, D).

Fig. 1: Donor-dependent variations of huMSC proliferation and immune modulation.
Eradication of specific donor-dependent variations of mesenchymal stem cells in immunomodulation to enhance therapeutic values

A Three lineage differentiation of huMSCs and expression of cell surface markers. Upper Panel from left to right: MSCs differentiated to adipocytes, chondrocytes, and osteocytes. B BV2-suppression capacity and proliferation capacity of 32 lines of huMSCs. The black bars represent the ratio of BV2 cell mass after 48 h cultured in MSC-CM compared to RPMI-1640 (control medium). The red dots represent the ratio of total huMSC mass at 72 h after inoculation compared to 1 h after inoculation. C correlation between BV2-suppression capacity and huMSC proliferation capacity of 32 huMSC lines. D Boxplot shows BV2-suppression capacity between male (n = 20) and female (n = 20) huMSC lines, eight extra huMSC lines were added to increase the statistical power.

Donor-dependent variations of huMSC immunomodulation in vivo

To investigate the effect of donor-dependent differences on potential clinical efficacy of MSCs, we selected two lines of huMSC with immune SI equal to 0.35 (MSC1) and 0.67 (MSC2), respectively. We established a mouse model of intra-peritoneal injection of LPS-induced neuroinflammation25 (Fig. 2A). After daily injection of LPS at a dose of 1 mg/kg, mice started to develop a series of abnormalities including lethargy (sleepiness and immobility), arched back (huddling), hair bristling (piloerection), ptosis, weight loss, and anal-rectal dysfunction. On the 4th and 6th days after the first LPS injection, mice were treated with CM from MSC1 and MSC2 via tail vein injections (Fig. 2A). Open field tests were performed on the 11th day to access neural deficits. Immobility after LPS treatment was obvious and could be quantitatively reflected by reduced ambulatory distance and episodes (Fig. 2B, C). Neural inflammation could be reflected by increased hippocampal TNF-α mRNA levels (Fig. 2D). Tail vein injection of CM from MSC2 (SI = 0.67) reduced neural inflammation and improved animal motor behavior (Fig. 2B–D), whereas CM from MSC1 (SI = 0.35) only showed weaker or a trend (without statistical significance) of improvement (Fig. 2B–D). Similarly, in a mouse crush injury-based spinal cord injury model, tail vein transfusion of huMSC-A with SI = 0.67 at a dose of 1 million cells/animal (weighed about 25 g) at day three post-surgery also improved animals’ walking behavior (by BMS scoring) (Supplementary Fig. S3). However, huMSC-C and B lines with SI = 0.26 and 0.52, respectively, did not significantly improve animal behavior (Supplementary Fig. S3). Together, these data suggested that donor-dependent variations of huMSCs in immunomodulation would likely affect their therapeutic efficacies, when treating neural-inflammation-related clinical conditions.

Fig. 2: Donor-dependent variations in immune modulation of huMSCs in LPS-induced neural inflammation animal model in vivo.
Eradication of specific donor-dependent variations of mesenchymal stem cells in immunomodulation to enhance therapeutic values

A Schematic diagram shows in vivo assessment MSC-CM treatment efficacies in LPS-induced neuroinflammation animal model. Open-field-test was used as the main behavioral readout. Boxplots show ambulatory distance (B) and episodes (C) of the open-field-test of each treatment groups. All data were normalized to the average value of the LPS group. D Boxplot shows inflammatory factors TNF-α mRNA levels in hippocampus of each group. Data were also normalized to the average value of the LPS group, and presented as mean ± SEM, *P < 0.05, **P < 0.01, ****P < 0.0001.

Transcriptomic analysis of huMSC showed positive correlations of IFN-γ and NF-κB signaling with immune modulatory function

To reveal molecular mechanisms underlying the immune suppressive function of huMSCs, we analyzed the transcriptome of all 32 lines of MSCs. Two-Dimensional Principle Component Analysis (PCA) showed MSC whole transcriptomes were dispersed, and not correlated well with BV2-suppressive function (Fig. 3A). We later isolated 1037 genes that show significant positive and negative correlations (784 positively correlated genes and 253 negatively correlated genes) with BV2 inhibition levels (Fig. 3B).

Fig. 3: Transcriptomic analysis of 32 lines of huMSCs revealed functional modules related to BV2-suppression capacity.
Eradication of specific donor-dependent variations of mesenchymal stem cells in immunomodulation to enhance therapeutic values

A PCA of 32 huMSC lines. B Heatmap shows expression and clustering of genes, expression of which significantly correlated to BV2 inhibition capacity of 32 lines of huMSCs. C GO enrichment analysis of BV2 inhibition correlated genes. IFN-γ, autophagy, and related terms are highlighted in red, which are positively correlated with BV2 inhibition. Candidate genes belonging to each GO term are shown in blue. D GSEA analysis of enriched TFBS targeting gene sets, most of which, except for one, were positively correlated with BV2 inhibition (normalized enrichment score (NES) > 0). E Correlation between gene expression levels of two subunits of NF-κB complex and BV2 inhibition capacity. F Network shows connection between enriched GO terms, TFBS targeting gene sets, and positively correlated genes. Red squares represent GO terms and blue triangles represent TFBS targeting gene sets. The nodes of green balls are genes, which belonging to both corresponding GO terms and TFBS targeting gene sets.

Gene ontology (GO) enrichment analysis showed that genes positively correlated with BV2 inhibition levels were linked to functions including “autophagy”, “interferon-γ mediated signaling pathway” and “histone deacetylation”, meanwhile genes negatively correlated with BV2 inhibition levels were related to “response to topologically incorrect protein” and “intermediate filament organization” (Fig. 3C and Supplementary Table S3). A network analysis based on GO semantic similarity also supported these results (Supplementary Fig. S4). Additionally, we analyzed transcription factor binding site (TFBS)-enriched gene sets in our data, results showed that both NF-κB related TFBS (NFKB_C, NFKB_Q6_01, NFKAPPAB_01) and interferon response TFBS (ISRE_01, IRF_Q6) were present in genes that were positively correlated with BV2 inhibition (Fig. 3D), indicating that huMSC-intrinsic expression levels of NF-κB signaling pathway and IFN-γ signaling pathway reflected immunomodulation capacities of huMSCs. Indeed, NF-κB subunits RELB and NFKB2 showed positive correlations with BV2 inhibition levels (Fig. 3E).

Further construction of a regulation network using genes positively correlated with BV2-cell inhibition, together with their GO terms and TF-binding, illustrated a more comprehensive network related to immunomodulation capacity of MSCs. The gene regulation network revealed many NF-κB pathway related genes, such as NFKB2, RELB, IKBKE, as well as IFN-γ signaling pathway genes. In fact, it has been reported that without stimulation of IFN-γ, MSCs showed very limited immunomodulation effect, and blockade of IFN-γ or IFN-γR impaired immunomodulation function of MSCs26,27. Interestingly, histone deacetylation as well as autophagy, the other two GO themes linked to BV2-cell inhibition, appeared to be also well connected to NF-κB and IFN-γ signaling (Fig. 3F)

Pretreatment with IFN-γ and TNF-α eliminated donor-dependent variations of huMSC in immunomodulation

Transcriptomic analysis suggested that differences in the expression of genes related to the IFN-γ and NF-κB signaling pathway, which could be activated by TNF-α, may be a critical reason for donor-dependent variations in huMSCs regarding immune modulation. It has also been well-known that inflammatory factors such as IFN-γ and TNF-α could enhance the immunomodulatory capacity of MSCs, and several studies have reported improvement of treatment efficacies of MSCs by pretreatment with inflammatory factors28,29,30.

We therefore decided to use the two factors (IFN-γ and TNF-α) to stimulate two MSC lines with SI equaled to 0.67 and 0.26, respectively, to determine whether line to line variations could be eradicated. As described in our previous work31, key anti-inflammatory factors IDO1, CXCL9, and IL6 were extensively upregulated after 2-factor stimulation (Fig. 4A). BV2-cell inhibition assays showed that the inhibition capacities of MSC1 and MSC2 were significantly increased after 2-factor stimulation and differences between the two lines were eliminated (Fig. 4B). The in vivo treatment efficacies of the two MSC lines were then examined, before and after stimulation, using the LPS-induced neural inflammation mouse model. As expected, MSC2 showed better therapeutic effect than MSC1 without 2-factor stimulation. However, after stimulation, the therapeutic efficacy of MSC1 was no longer statistically different from that of MSC2 (Fig. 4C–E). These results suggest that 2-factor stimulation may improve the anti-inflammatory function of huMSCs, and eradicate the donor-dependent variations.

Fig. 4: Two-factor stimulation abolished donor-dependent variations of huMSCs in immune suppression.
Eradication of specific donor-dependent variations of mesenchymal stem cells in immunomodulation to enhance therapeutic values

A Anti-inflammation genes IDO1, CXCL9 and IL6 were extensively up-regulated in huMSCs by stimulation of TNF-α and IFN-γ (2-factors). B In vitro analysis on BV2 inhibition showed elimination of donor-dependent variations in immune-modulations by two factors. CE eradication of specific donor-dependent variations of huMSCs regarding treatment efficacies in LPS-induced neuroinflammation model, by open-field-motor scoring and hippocampal TNF-α expression. All data were normalized to LPS group, and were presented as mean ± SEM, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

IFN-γ and TNF-α stimulation of huMSCs enhanced immunomodulation but suppressed proliferation

To reveal whether inflammatory factor stimulation could mobilize the intrinsic regulatory network of MSCs to enable global enhancement of immunomodulation, and thus make the different huMSC lines overall more similar, we analyzed the transcriptomes of huMSCs before and after stimulation. PCA results showed that inflammatory factor stimulation dramatically altered the transcriptomes of huMSCs (Fig. 5A), but the three different huMSC lines were still different at whole transcriptomic levels. We identified 9702 differentially expressed genes (4724 upregulated and 4978 downregulated) before and after stimulation (Fig. 5B). Interestingly, the functional enrichment analysis showed that among the genes that were significantly upregulated after stimulation, the greatest enrichment was in DNA replication and cell cycle-related functions (Fig. 5C and Supplementary Table S4). A network analysis based on the semantic similarity of GO terms revealed that genes that were upregulated after stimulation were concentrated in five functional modules: DNA replication and cell cycle, response to IFN-γ and TNF-α signals, antigen processing and presentation, macromolecular translocation, macromolecular synthesis, and degradation (Supplementary Fig. S5). On the other hand, there were four functional modules enriched in genes downregulated by 2-factors: TGF-β and SMAD signaling pathways, cell differentiation, vesicle assembly, and macromolecular metabolic process (Supplementary Fig. S6).

Fig. 5: Transcriptome analysis of huMSCs before and after 2-factor treatment.
Eradication of specific donor-dependent variations of mesenchymal stem cells in immunomodulation to enhance therapeutic values

A PCA of all samples including 3 different genetic background and 2 conditions. B Heatmap demonstrating differentially expressed genes (DEGs) between 2 conditions. C GO enrichment analysis of DEGs in B. D Bar-plot demonstrating percentage of proliferating cells before and after 2-factor stimulation based on EdU assay. E Bar-plot shows percentage of cells in each cycle phase before and after cytokine stimulation based on flow cytometry.

To verify the results of the transcriptome analysis, we examined the proliferative capacity of huMSCs before and after stimulation and found a significant decrease in the number of proliferating cells after 2-factor stimulation (Fig. 5D). In addition, cell cycle analysis indicated that inflammatory factor stimulation retained huMSCs in the G1 phase and consequently reduced the number of cells in the G2 phase (Fig. 5E). This confirmed the transcriptomic results.

We further analyzed the 227 genes, which were both upregulated by 2-factor stimulation and were positively correlated with capacities of BV2-cell inhibition (Fig. 6A, B). Function of these genes focused on immune related pathways included IFN-γ signaling pathway, Type I interferon production and antigen processing. Interestingly, based on gene set variation analysis (GSVA), genes linked to “protein deacetylation” and “autophagy”, while positively correlated with BV2 immune suppressive capacities of MSCs, were not upregulated upon 2-factor treatment (Fig. 6C and Supplementary Table S5). These results suggested that 2-factor stimulation, while potentially enhancing the immunomodulatory function of huMSCs by activating IFN-γ and NF-κB signaling pathways and eradicating donor-dependent variations of huMSCs, also made huMSCs proliferate much slower. Whether or not this is a desired feature for MSC-based therapeutic interventions would depend on the detailed therapeutic targets/purposes.

Fig. 6: Two transcriptomic data sets (32 huMSC lines and 3 huMSC lines with and without 2-factor stimulation) revealed common genes and pathways.
Eradication of specific donor-dependent variations of mesenchymal stem cells in immunomodulation to enhance therapeutic values

A Venn diagram shows the overlap between genes up-regulated by 2-factor stimulation and genes positively correlated to BV2 inhibition capacity. B GO enrichment analysis of overlapped 227 genes. C GSVA heatmap shows specific GO-term associated gene sets in 32 huMSCs that positively or negatively correlated with BV2-cell immune suppressive capacity, and their expression levels before and after 2-factor stimulation. Results clearly indicated that interferon and NF-κB pathway associated genes were upregulated upon 2-factor stimulation, whereas genes related to autophagy and protein deacetylation did not correlate well with 2-factor stimulation.

Read original article here.