APS processing enriches immune concentrations of multiple immune cell populations
We characterized the immune cell populations after APS processing in blood acquired from 4 healthy human donors, Fig. 1a. We used flow cytometry to investigate the immune cell types present in APS; we defined populations of neutrophils (CD45+CD3−CD15+CD16+), monocytes (CD45+CD3−CD15−CD14+CD16dim), natural killer cells (NK) cells (CD45+CD3−CD15−CD14−CD16+HLA−), NKT cells (CD45+CD3+CD14−CD16+), and T cells (CD45+CD3+CD14−CD16−) based on cell surface markers, gating strategy shown in Fig. S1a. APS processing resulted in a significant 5-fold enrichment of CD45+ immune cells while maintaining cell viability when compared to WBCs (Fig. 1b). Neutrophils (24 million ± 11 million cells/mL) and T cells (9.8 million ± 6.9 million cells/mL) were the most abundant immune cell types in APS followed by classical monocytes (1.9 million ± 1 million cells/mL), NK cells (1.6 million ± 0.80 million cells/mL), and NKT cells (1.4 million ± 1.1 million cells/mL) (Fig. 2c). Neutrophils, classical monocytes, and NK cells were significantly enriched in APS compared to WBCs. Other immune cell types were enriched after processing and present at lower concentrations including non-classical monocytes (CD45+CD3−CD15−CD14dimCD16+), dendritic cells (CD45+CD3−CD15−CD14-HLA+CD16−CD11c+HLA+), and eosinophils (CD45+CD3−CD15+CD16−) (Fig. S2a-b). Eosinophils experienced significant enrichment after APS processing.
We then explored the donor variability. Donor samples were processed on the same day, and the flow cytometry experiment was performed on the same cytometer at the same settings. The variation among the same immune cell types between donors shown in the forward and side scatter plots suggests donor differences (Fig. 1a). There are apparent shifts in the subpopulations between individual donors. The classical monocytes (CD45+CD3−CD15−CD14+CD16dim, gray) are more abundant in APS relative to WBCs and shift among different donors in the forward and side scatter plots. The neutrophils (CD45+CD3−CD15+CD16+, black), in particular, have large variability between volunteers in terms of size and granularity. Shifts among granulocytes due to density changes during inflammation suggest differences in activation status, maturation state, and granulation [17, 18]. Overall, after APS processing, monocytes (16 times more abundant), followed by neutrophils (6 times more abundant), had the largest fold change among the immune cell populations when compared to WBCs (Fig. 1d).
Distinct differential gene expression changes in immune cell subtypes after APS processing is immune cell type-dependent
We sorted granulocytes (CD45+CD11b+CD3−CD15+CD14−), T cells (CD45+CD3+CD11b), and monocytes (CD45+CD3−CD14+CD15−CD11b+CD11c+) from APS and WBCs to evaluate the mechanistic pathway activity (FACS gating strategy shown in Fig. S3). To assess the gene expression changes related to the innate and adaptive immune response, we performed the Nanostring multiplex gene expression assay using the PanCancer Immune Profiling CodeSet. Data were analyzed using the ROSALIND platform (padj < 0.05, 1.5 ≤ LFC ≤ − 1.5), and analysis revealed differential changes in APS compared to WBCs dependent on cell type. The ROSALIND platform is a cloud-based software platform for gene expression data analysis. With respect to the genes analyzed, T cells remained largely unaffected after APS processing with no significant differentially expressed genes when compared to T cells sorted from WBCs. The myeloid compartment experienced significant changes after APS processing; 109 and 4 genes were differentially regulated from sorted monocytes and granulocytes, respectively (Fig. 1e). The differentially regulated genes in sorted granulocytes were related to antigen processing (Tap1) and inflammatory responses (Mapkapk2, Ptgs2) (Fig. S2c). We used STRING to construct a protein-protein interaction (PPI) network of the differentially regulated genes. Using the KEGG Pathways database we found functional enrichment of the VEGF signaling pathway (Mapkapk2, Ptgs2) and c-type lectin receptor (CLR) signaling pathway (Mapkapk2, Ptgs2). VEGF signaling has an important role in leukocyte recruitment and is involved in the wound healing cascade by promoting endothelial cell proliferation and angiogenesis, while CLRs can control adaptive immunity. The differences in the gene expression, particularly those found in monocytes after APS processing, may suggest functional changes that could impact treatment outcomes. Notably, all differentially expressed genes from sorted immune cells were all upregulated, suggesting a polarized response after APS processing.
To explore the impact of APS processing on the immune cell types, we created different visualizations of the Nanostring data. The sample correlation heatmap of the Nanostring data highlights that regardless of APS processing, sorted T cells and granulocytes from both WBCs and APS have a strong correlation, while monocytes sorted from APS or WBCs have a low correlation (Fig. S2d). The sample correlation heatmap is a graphical representation of the data, where individual correlation values (represented as colors) are contained in a matrix with dark red corresponding to strong correlation (close to 1). This trend is further supported by the multidimensional scaling (MDS) plot of the expression differences between the sorted cell types, where T cells and granulocytes from APS and WBCs cluster closer together suggesting similarities in the gene expression, while monocytes from APS and WBCs do not cluster together (Fig. S2e). Coordinate 1 of the MDS plot is distinguished by cell type, while coordinate 2 stratifies samples according to the processing method. Of the 770 genes assayed, these data suggest that the APS processing alters monocyte gene expression more than T cells or granulocytes. This alteration may have implications on the monocyte activity during APS treatment.
APS processing enriches M2-like dominant phenotype in monocytes
We investigated the monocyte phenotype after APS processing with flow cytometry. Here, changes in monocyte phenotype after APS processing were first approximated using macrophage polarization markers for M1- (CD80) and M2-like (CD163) phenotype, although the simplification of this classification does not address the heterogeneity and complexity that may exist. The mean fluorescence intensity (MFI) values for CD80 and CD163 were comparable among WBCs and APS samples. MFI values for CD163 were higher than MFI for CD80, suggesting an M2-like skewed phenotype for monocytes in WBCs prior to APS processing (Fig. 2a). MFI values for MHC-II were also comparable among WBCs and APS samples. APS processing resulted in a significant enrichment in CD163+ population in both classical monocytes from 158,000 ± 113,000 cells/mL to 1.74 million ± 1 million cells/mL. After APS processing, non-classical monocyte CD163+ population increased from 5200 ± 3000 cells/mL to 68,700 ± 66,000 cells/mL. Double-negative, CD80−CD163− nonclassical monocytes were enriched from 10,300 ± 6500 cells/mL to 102,000 ± 114,000 cells/mL (Fig. 2b). CD163+ expression is induced by anti-inflammatory mediators and is important in resolving inflammation. The expression of CD163 may also suggest homeostatic, regulatory, or immune maintenance programming [19, 20]. Other monocyte subpopulations were present in smaller concentrations and were not significantly enriched after APS processing (Fig. S4a).
To further characterize the monocytes, we evaluated the percentage of CD163+ and CD80+ in classical (CD14+CD16dim) and non-classical monocytes (CD14dimCD16+). CD163+CD80− was the dominant subtype in classical monocytes from APS and WBCs, 93.3% and 92.3%, respectively, while most non-classical monocytes were double-negative making up 58% and 66% of the population in APS and blood, respectively (Fig. 2c). There were few double-positive CD80+CD163+ and CD80+CD163− classical monocytes and non-classical monocytes. APS processing resulted in a 16.5× ± 13.7 enrichment of a M2-like phenotype for classical monocytes (CD163+CD80−), while other subpopulations were enriched at lower amounts such as double negative CD80−CD163− at 9× ± 6.4 or double-positive CD163+CD80+ at 1.1× ± 2.2, similar enrichment occurred in nonclassical monocytes (Fig. S4b). These findings suggest that APS processing further enriched the M2 polarization that was present among the donors and processing itself did not induce a phenotypic change. APS enriches CD163 in classical monocytes and non-classical monocytes by 17 and 12 times, respectively, the amount found in WBCs (Fig. 2c). To further characterize the monocytes, we evaluated the expression of antigen presentation marker, MHC-II. MHC-II+ monocytes 16% and 14% of classical monocytes in APS and in WBCs, respectively, while making up a larger percentage of non-classical monocytes, at 40% and 35% in APS and WBCS, respectively (Fig. 2d).
GO analysis, Nanostring annotations, GSEA, and PPI network analysis of gene expression data show upregulation of antigen presentation and processing pathways in sorted monocytes
Further profiling analysis of the sorted monocyte (CD45+CD3−CD14+CD15−CD11b+CD11c+) population from APS and WBCs using the multiplex gene expression assay from Nanostring performed using the Rosalind platform found that there were 109 differentially regulated genes when comparing monocytes sorted from APS to monocytes sorted from WBCs; the top 50 differentially regulated genes are shown in Fig. 2e. To explore the predicted functional associations of the differentially expressed genes, a STRING protein-protein interaction (PPI) network was constructed only from the proteins that have physical interactions such as proteins that are part of a physical complex and visualized using Cytoscape 3.9.0, an open-source platform for complex network analysis in Fig. S4c [21]. The PPI network is highly connected with 107 nodes (differentially expressed genes) and 268 edges (interactions between nodes). We then used tools for knowledge discovery to predict important genes in the network for additional insights into how gene expression changes due to APS processing could impact biological processes or pathways. To determine essential genes in the network, centrality measures were quantified using CytoNCA 2.1 [22]. Based on the top-ranked centrality measures for subgraph centrality, betweenness centrality, and closeness centrality, essential genes in the network are Fyn, Jak1, Jak3, Mapk1, Stat1, Stat3, and Syk. These genes are represented in green in the PPI network in Fig. S4c.
To identify highly interconnected regions in the PPI network nodes based on the connectivity degree of networks, the MCODE, molecular complex detection algorithm Cytoscape plugin was used [23]. The module analysis identified the most significant module containing 19 nodes, Jak1, Jak3, Stat3, Stat1, Fyn, Stat6, Stat5b, Il4r, Il6r, Il6st, Hla-dra, Hla-c, Hla-dpb1, Hla-g, Hla-b, Hla-a, Hla-e, Tapbp, Tap1, with 69 edges. The subnetwork of this module is represented in orange with the first neighbors represented in blue as depicted in Fig. S4d. The essential genes identified by the centrality measures were all in the significant module. Notably, 12 of the 19 are in the top 50 differentially expressed genes.
To further explore potential changes in the biological function and key pathways, Gene Ontology (GO) analysis was performed using STRING. Gene Ontology (GO) analysis of all significantly differentially regulated genes found that for the category of biological processes, sorted monocytes from APS had functional enrichment related to the following GO terms relevant to tissue repair: “IL-4-mediated signaling pathway and antigen processing” and “presentation of endogenous peptide via MHC class I via ER pathway, TAP-independent.” The molecular function category found enrichment in terms related to antigen processing such as TAP binding and TAP2 binding. The cellular component category found enrichment in macrophage migration inhibitory factor receptor complex. Functional enrichment using the protein domains (Pfam) database found enrichment in the following domains: MHC-I C-terminus and STAT protein (protein interaction domain, DNA-binding domain, and all alpha-domain).
Additional pathway collections were explored for enrichment analysis using the Rosalind platform (padj < 0.05). MySigDB Pathway Collection found several significantly enriched terms associated with antigen processing and signaling processes (Fig. 2f). Enrichment was found for the following MSigDB REACTOME terms, Class I MHC mediated antigen processing presentation (Ubc, Psmb8, Tap1), antigen processing cross-presentation (Psmd7, Ly96, Hla-g), antigen presentation folding assembly and peptide loading of class I MHC (Hla-g, Hla-c, Hla-e), and adaptive immune system (Sell, Psmd7, Ly96). The Wiki pathways showed enrichment of proteosome degradation, IL-4 signaling pathway (Fos, Stat3, Tyk2), IL-2 signaling pathway (Stat1, Fyn, Syk), and oncostatin M signaling pathway (Prkcd, Jak1, Jak3).
We then performed gene set enrichment analysis using Nanostring Annotations on the Rosalind platform and found enrichment of terms related to cytotoxicity (Hla-a, Hla-b, Hla-b), antigen processing (Tap1, Tapbp, Psmb9, Psmb7), transporter functions (Fyn, Itgam, Cd47, Cd44), and cell cycle (Cxcr4, Tnfsf10). The overall differential expression of each gene set (directed enrichment score) showed 4 times differential expression in APS sorted monocytes compared to blood (Fig. 2g).
Altogether, the pathway analyses identified antigen processing-associated gene sets or terms in all the databases and collections used. Tap2, an antagonist peptide of TLR-4, had analgesic and anti-inflammatory effects in a monoiodoacetate (MIA)-induced rat model of OA that resulted in decreased cartilage loss [24]. Tap2 can reduce ROS in the arthritic joint. Pathways and terms identified from the analyses employed suggest an association with anti-inflammatory functions in monocytes (Ifi16, Stat3, Il10ra, Cxcr4, Cd63, Cd47). Stat3, Il10ra, and Cxcr4 were in or the first neighbors of the densest region of the PPI network and had more than 4 times fold change in expression.
Lymphocytes are enriched and maintain phenotypic subsets after APS processing
We used flow cytometry to investigate the impact of APS processing on lymphocyte concentration. CD3+ T cells were more prevalent than CD3−CD19+ B cells. APS processing significantly increased the concentration of T cells and B cells from 1,710,000 ± 547,000 cells/mL and 310,000 ± 40,000 cells/mL in WBCs to 11 million ± 3,800,000 cells/mL and 1 million ± 384,000 cells/mL after APS processing, respectively (Fig. 3a). CD4+ T cells were the dominant T cell subtype, then CD8+. CD4+ and CD8+ T cell subsets were significantly increased from 1,210,000 ± 420,000 to 8 million ± 3,300,000 cells/mL and 343,000 ± 143,000 to 2,430,000 ± 1,380,000 cells/mL, respectively, after APS processing (Fig. 3b).
We then used intracellular staining to investigate phenotypes of T cells. After APS processing, CD4+ and CD8+ T cells maintain the potential to express cytokines as shown by intracellular staining after stimulation with PMA/ionomycin, gating strategy shown in Fig. S5. CD4+ IFNγ+, IL-4+, and IL-17A+ were significantly enriched after APS processing compared to WBCs. CD4+IFNγ+ T cells were enriched from 303,000 ± 297,000 to 1.42 million ± 582,000 cells/mL, CD4+IL-4+ T cells increased from 24,000 ± 17,000 to 99,000 ± 36,000 cells/mL, CD4+IL-17A+ T cells 19,000 ± 15,000 to 74,000 ± 18,000 cells/mL (Fig. 3c). CD8+IFNγ+ T cells were also significantly enriched after APS processing, with an increase from 184,000 ± 115,000 to 990,000 ± 340,000 (Fig. 3d). These findings suggest that the functionality of the WBCs is maintained after APS processing and that the second most abundant immune cell subpopulations in APS, T cells, could be an additional source of bioactive factors.
Immune cell fraction persists after APS treatment in RAG KO OA model
To explore how different components of human-derived APS can impact local response to an injury, we used a post-traumatic osteoarthritis model in 10-week female C57BL/6 RAG KO mice. Two weeks after inducing unilateral OA via ligament transection, experimental treatment groups (n = 5/group) were injected with either whole human APS, 3000 sorted CD45+ cells from APS, 3000 sorted CD3+ cells from APS, or the acellular portion of APS (containing soluble proteins only). Saline injections and no surgery groups were used as controls. Injections of 20 μL for each experimental group ensured the maintenance of the integrity of the joint capsule and localization of the APS near the joint (study design in Fig. 4a). Two weeks after injections, for each experimental group, the animals were harvested for histology of the joints (n = 1/group) and qRT-PCR of the inguinal lymph nodes (n = 5/group) and the remainder of the joints (n = 4/group). Weight-bearing measurements were not restored for any of the treatment groups when compared to no surgery controls. Significantly longer response times in the APS-treated group suggest that the group could be experiencing more pain than the saline and APS without cell groups (Fig. 4b). Representative safranin-O staining (n = 1/group) is provided in Fig. 4c to observe proteoglycan content in the joint.
We then investigated the survival of the cell fraction of APS after injections. Immunofluorescence staining confirmed the persistence of the cell fraction of APS 2 weeks after single injections. The presence of human positive CD4 cells in the synovium of the APS treatment groups with cells (APS, CD3+ sorted from APS, and CD45+ sorted from APS) demonstrates cell persistence and the durability of the cell fraction that may be important for APS treatment outcomes (Fig. 4d). Tissue-resident immune cells can be long-lived. Lymphocytes can persist for weeks to even years as is the case for memory T cells. The injected immune cells persist beyond the lifespan of platelets contributing to the microenvironment well beyond the initial injection and the biological activity of proteins in APS.
To investigate the potential systemic effects of the treatment, the gene expression of pro-inflammatory and anti-inflammatory cytokines in the inguinal lymph node that drains lymphatics from the injured articular joint were evaluated. Il1b, Il10, Ifng, and Il4 did not significantly change among the experimental groups in the draining inguinal lymph nodes suggesting that the injections did not polarize the systemic response (Fig. S7a). IL-17 cytokines have demonstrated pathophysiology in osteoarthritis in preclinical and clinical studies [25]. Quantification of Il17f mRNA in the draining inguinal lymph node increased in some mice in the saline and APS without cell treatment groups compared to the other groups, although did not reach statistical significance (Fig. S7a). Il17a was not detected in the joint or lymph node for all groups.
To investigate the local effects of the treatment, the gene expression of relevant genes of the whole joint were explored. A broad-spectrum MMP inhibitor important for reducing cartilage degradation, Timp1, had significantly increased expression in the joint for all treatment groups compared to no surgery control. Regulatory cytokine, Tgfb1, had significantly increased expression in the joint for APS, APS without cell, and CD3+ sorted from cells from APS treatment groups compared to no surgery control. Tgfb2 expression was not significantly differentially expressed among treatment groups. Il17b expression in the joint did not significantly change among the treatment groups. Additionally, the pro-inflammatory cytokine expression in the joint, Tnfa, metalloproteinase Mmp12, and Ddr1, known to control MMP-13 expression during chondrogenesis, had no significant difference among the groups (Fig. S7b). Pro-inflammatory cytokine Il1b expression was significantly increased for the CD3+ sorted from the APS group when compared to all groups.
The cell fraction of APS is the main source of anti-inflammatory IL-1Ra cytokine and is enriched after APS processing
APS can be prepared from a broad range of OA patients with consistent preferential enrichment of anti-inflammatory cytokines [4]. The pleiotropic nature and kinetics of cytokines in the wound healing cascade require further understanding of the OA pathology and treatment. To evaluate the concentration of prominent cytokine and growth factors studied in the wound healing cascade, we performed Luminex cytokine assays on the cellular compartment (lysates) and the soluble fraction (plasma) of APS and blood. All white blood cell lineages produce cytokines. Once activated, platelets release stored cytokines and chemokines. We did not distinguish the Luminex results by cell type. Based on the literature, we can hypothesize what cell types may be contributing to the source of relevant bioactive factors. In APS, IL-1Ra was the most abundant protein (103,000 ± 50,000 pg/mL), followed by IGF-1 (93,000 ± 17,000 pg/mL), and TGF-β1(15,000 ± 8000 pg/mL). IGF-1 is a growth hormone mainly produced by the liver and then released into circulation [26]. TGF-β1 is a pleiotropic cytokine produced by all white blood cells [27]. IL-1β and TNF-α are known pro-inflammatory cytokines that promote OA pathophysiology and are mainly produced by macrophages [28]. IL-1β and TNF-α can also be released by activated platelets [29]. TNF-α is also secreted by other immune cells such as monocytes, T cells, NK cells, and NKT cells [30]. IL-1β and TNF-α were present at low levels, 57 ± 5 pg/mL and 40 ± 27 pg/mL, respectively, in APS (Fig. S6a). Other studies have noted the presence of remarkably low concentrations of IL-1β in APS [7]. Another study characterizing the cytokine profile of APS did not detect TNF-α [4]. The non-cellular soluble protein fraction significantly enriched other anabolic cytokines such as TGF-β1 and IGF-1 by 5× and 2×, respectively (Fig. S6b). The balance between IL-1Ra and IL-1β has a critical role in the homeostasis of inflammation. It is important to note the concentrations of the cytokines present. Although pro-inflammatory cytokines like IL-1β and TNF-α are present in APS, anti-inflammatory cytokines are present at concentrations 4–5 orders of magnitude higher.
Among anti-inflammatory cytokines enriched in APS, the assay revealed a significant increase, 17× fold change, of IL-1 receptor antagonist (IL-1Ra) specific to the cellular fraction (lysates). The significant increase in IL-1Ra is comparable to other studies characterizing APS cytokine enrichment [4]. IL-1Ra was not detectible in the plasma fraction. This is notable because previous studies have found that APS has a 5.9-fold increase in IL-1Ra and 5-fold increase in WBCs compared to whole blood [4]. Findings from the first in-human clinical trial with a single injection of APS found that subjects with high concentrations of WBCs and IL-1Ra:IL1-β ratio greater than 1000 in APS were more likely to respond to the APS therapy than the total study population [31]. WBC concentration was significantly correlated with IL-1Ra in APS, as predicted given that the common isoforms of IL-1Ra are generally found inside of the cell [32]. The intracellular isoform of IL-1Ra is primarily produced by neutrophils and monocytes [33,34,35]. Neutrophils are the dominant immune cell type in the final product of APS and are likely the main source of potent anti-inflammatory intracellular IL-1Ra. The high IL-1Ra:IL1-β ratio for the healthy volunteers in the present study suggests that they would respond to APS injections. The Luminex cytokine assay confirmed that the cell fraction of APS is the major source of IL-1Ra, as shown in previous studies [7]. Of the analytes studied, IL-1Ra was the only analyte detected in the cell fraction. IGF-1, IL-1β, TGF-β1, TGF-β2, and TGF-β3 were not detected in the cell fraction of APS. TGF-β3 was not detected in either APS or blood. A small concentration of TNF-α (4.7 pg/mL) was detected in one of the APS cell fraction samples. Notably, TGF-β2 (386 ± 38 pg/mL) was detected in the blood plasma fraction but not in the plasma or cell fraction of APS (Fig. S6a).