Restoration of aged hematopoietic cells by their young counterparts through instructive microvesicles release

Hematopoietic restoration of aged mobilized peripheral blood cells (MPBs)

MPBs are comprised of mixed cell subsets that include mature and immature hematopoietic cells. Furthermore, although the cellular frequency of particular cell subsets are different between young and aged MPBs, the compositions are similar. Thus, we evaluated these two sources of MPBs for baseline hematopoietic cells in clonogenic assays for CFU-GM (colony forming units-granulocytic monocytic) and BFU-E (burst forming units-early erythroid). There were similar numbers of CFU-GM between the two groups, but significantly (p < 0.05) less BFU-E in young MPBs (Supplementary Figure 1A). Flow cytometric analyses indicated comparable frequency of CD34+/CD38- cells (primitive hematopoietic cells) in the young and aged MPBs, but significantly (p < 0.05) more CD34+/CD38+ (progenitors) in young MPBs (Supplementary Figure 1B).

The hematopoietic effects by young MPBs on aged MPBs were studied in a transwell system in which the inner and outer wells contained equal amounts (10⁷) of young and aged MPBs, respectively (heterochronic culture) (Figure 1A). Control isochronic cultures contained aged or young MPBs in both wells. The 0.4 μm transwell membrane allowed for the passage of soluble factors and microvesicles (MVs), but prevented cell transfer [37] (Supplementary Figure 1C). The initial studies conducted a time-course investigation up to wk 5 using clonogenic assays for CFU-GM with the aged cells as readouts of hematopoietic activity. CFU-GM progenitors are excellent representation of hematopoietic activity in the samples. Controls included isochronic cultures with aged or young MPBs in both inner and outer wells. There was significantly (p < 0.05) more CFU-GMs with heterochronic cultures as compared to isochronic aged cultures but similar numbers (p > 0.05) with isochronic young cultures (Figure 1B). We also evaluated the phenotype of the restored aged cells and observed mature immune cells such as CD3 and NK cells (Supplementary Figures 1 and 6). These timeline studies combined with the phenotype shown at wk 7 led us to select the 7-day time point for subsequent studies.

Figure 1. In vitro hematopoietic restoration of aged MPBs. (A) Cartoon shows the method employed for non-contact isochronic and heterochronic cultures. (B) Timeline clonogenic assays for CFU-GM with viable cells from isochronic (aged or young MPBs) and heterochronic cultures (restored aged MPBs). The results are presented as mean CFU-GM ± SD (n = 8 donors, each donor tested in triplicate with two young donors). ^*p < 0.05 vs. similar time points in heterochronic cultures; ^**p < 0.05 vs. time 0 and wks 2, 3 and 4. (C) LTC-IC cultures were established using the model in ‘A’ except for seeding the aged MPBs on confluent γ-irradiated BM stromal cells in the outer wells. Control cultures contained isochronic young or aged cells in both wells. At wks 1, 6 and 12, clonogenic assay for CFU-GM with aliquots of viable mononuclear cells. The values for each time point were plotted together (8 donors, each tested in triplicates, CFU-GM/10⁵ MPBs ± SD. ^*p < 0.05 vs. heterochronic. (D) Heterochronic and isochronic cultures were established with 10⁷ UCB in the inner wells. At wk 4, aliquots of aged MPBs were analyzed for CFU-GM and the results presented as mean CFU-GM ± SD for 5 different UCB, each tested in duplicate. ^***p < 0.05 vs. heterochronic cultures with UCB. (E) Senescence-related gene expression was performed with 84-gene qPCR arrays using cDNA from restored and unrestored (baseline) aged and young MPBs. Gene expression for 4 donors was determined by calculating the ΔC_t between gene-of-interest and housekeeping genes and then plotted as Log₁₀(2^ΔCt). Each dot represents the average gene expression for donors. Baseline comparison for unrestored young vs. aged MPBs is shown in red for higher expression in young and green for higher expression in aged. The line y = x indicates no change. (F) The analyses described in `E’ was performed for young and restored and the data are similarly presented. (G) Oxidative stress by MitoSox assay, delineated as MitoSox, negative, low and high by flow cytometry. (H) Hierarchical clustering with the array data from ‘E and F’. (I) The genes upregulated in the qPCR array in ‘H’ (open boxed region) were analyzed by RAIN to demonstrate predicted interactions. (J) Western blot (3 biological replicates) with whole cell extracts from unrestored young MPB (Y), UCB and restored A1 (restored with UCB or Y-MPB). SDS-PAGE: top, 15%; middle 12%; bottom, 6%. (K) PCA of RNA-Seq data from MPB (3 young, 3 age) and 4 restored MPBs. Lines highlight the groups. (L) Similarity matrix of ‘A’ for young, age and restored samples. (M) Volcano plot of differentially expressed genes. (N) Heatmap of fold changes with an FDR ≤ 0.05 as a cut off with linked significant pathways. (O) IPA-determined significant hematological functions with shown comparisons. See also Supplementary Table 1, Supplementary Figures 1 and 2.

The enhanced activity by hematopoietic progenitors (Figure 1B) led us to ask if similar outcome occurred at the level of HSCs. We performed long-term culture-initiating cell (LTC-IC) assay in the outer wells of heterochronic cultures since this assay is an in vitro surrogate of HSC function [38]. Control isochronic cultures included young or aged cells in the inner and outer wells. At wks 1, 6 and 12, CFU-GMs in the outer wells of heterochronic cultures (n = 8, each in triplicate) were significantly (p < 0.05) increased, relative to isochronic aged cultures but were similar to isochronic young cultures (Figure 1C). Since CFU-GMs at wk 12 are derived from the seeded HSCs, their increase indicated increase in LTC-IC/HSC function with young cells.

Hematopoietic restoration - independent of donor allogenicity

We determined if allogeneic differences between the age and young donors influenced hematopoietic activity in the heterochronic cultures. Such possibility could occur by MHC-II transfer from the young cells to activate the immune cells within the aged cells [39, 40] (Supplementary Figure 1D). We addressed this question by asking if the restored aged cells stimulated their naïve counterparts (unrestored autologous cells). One-way mixed lymphocyte reaction (MLR) with restore aged MPBs as stimulators and unrestored/freshly thawed aged autologous MPBs as responders, showed baseline stimulation, indicating no allogeneic influence by the young cells (Supplementary Figure 1E).

Next, we asked if MHC-II density on young cells could influence the restoration process. We used umbilical cord blood (UCB) as the source of young cells due to relatively higher MHC-II on their HSCs, which we corroborated in our samples [41] (Supplementary Figure 1F). Parallel heterochronic cultures with UCB or young MPBs and the same aged cells indicated significantly (p < 0.05) more CFU-GM with UCB, relative to isochronic aged cultures, but significantly (p < 0.05) less than young MPBs (Figure 1D). One-way MLR with UCB showed no allogeneic response (Supplementary Figure 1G).

MVs can transfer MHC-II leading us to examine these MVs for MHC-II [37, 42]. Flow cytometry for MHC-II on heterochronic MVs indicated background fluorescence whereas the positive control (activated peripheral blood mononuclear cells) showed bright fluorescence (Supplementary Figure 1H and 1I). Together, the data indicated that allogeneic difference between the young and aged cells did not contribute to aged cell restoration.

Improved senescence of aged cells in heterochronic cultures

Cellular senescence is an established aging hallmark, which prompted us to ask if the heterochronic cultures improved the senescence gene profile of these cells [10, 43]. Senescence gene arrays (84-PCR), using a 4-fold cutoff, indicated 85% increase for unrestored/baseline aged cells and 43% for restored cells (Figure 1E, 1F, Supplementary Figure 2A and 2B). The values were similar with variations for each time point, ±10³. Since the purpose of the studies was to evaluate the changes in senescence genes, we plotted the averages in the scatter plot. Similar analyses with senescence secretome arrays (SASP, 68 antibodies) identified 13 proteins in the media of baseline aged cells and 4 with in the media of heterochronic/restored cultures [44] (Supplementary Figure 2C–2G). Collectively, the restoration paradigm improved the senescence gene profile of aged MPBs.

Molecular changes - amenable to balanced hematopoietic activity

The vast difference in oxidative stress between young and aged samples (Figure 1G) led us to further analyze the senescence PCR data. Hierarchical clustering using the averages from biological replicates indicated striking similarities between young and restored MPBs (Figure 1H). A cluster of 30 expressed genes (Boxed region), when subjected to RNA-protein analysis (RAIN), showed a gene network comprised of hematopoietic regulatory programs (Figure 1I, Supplementary Figure 2H). Including is MDM2 expression that can decrease p53, which is consistent with improved senescence profile in the restored samples. Decreased p53 would not compromise DNA repair due to CHEK1/2 within the network [45]. Upregulated IGF1, which is associated with longevity, was increased in the restored cells [46]. Decreased fibronectin (FN1) in the restored cells could improve risk for hematopoietic pathology such as myelofibrosis [47].

We validated key proteins within the network (Figure 1I-underline) by Western blot using extracts from aged cells (Figure 1J, densitometric analyses, Supplementary Figure 2I). Increases in p-PTEN and p-AKT in the restored cells are amenable to balanced cell proliferation; increased cyclin A can decrease G2 transition to maintain hematopoietic cell quiescence; higher GSK-3β suggested a pathway involving β-catenin to benefit hematopoietic function. Consistent with increased MDM2 in the PCR array was decreased p53 protein (Supplementary Figure 2I). In summary, hematopoietic restoration is associated with decreased senescence and increased expression of genes to benefit hematopoietic activity.

Molecular landscape in unrestored versus age restored MPBs

We examined global changes with RNA-Seq by comparing the following: restored and unrestored age cells, and young MPBs. Principle component analyses (PCA) showed distinct clustering among the groups (Figure 1K). Although one young donor was not within the cluster of the other two young donors, similarity matrix showed the restored samples moving closer to the young profile (Figure 1L). Applying a false discovery rate (FDR) cutoff of <0.05, we selected 2,140 genes, visualized in volcano plots and hierarchical clustering (Figure 1M and 1N). Since the hierarchical plot was established with >2,000 genes, we conducted additional analyses to understand the molecular changes with the restored aged MPBs.

Ingenuity pathway analyses (IPA) identified top pathways associated with hematopoietic development and, decreases in reactive oxygen species, apoptosis and necrosis (Figure 1O). Gene Set Enrichment Analysis (GSEA) identified significantly up- and down-regulated (FDR q value < 0.05) pathways in age and young groups (Figure 2A and 2B). MYC pathways were the only downregulated targets in the age group with 14 upregulated pathways. Restoration improved key age-related functions - inflammation, immune suppression and cell death - TNFα signaling, NFκB signaling, apoptosis and p53. The Venn diagram depicts the pathways corrected by restoration along with the shared and unique pathways (Figure 2C). The up- and down-regulated pathways in the aged samples transitioned into the overlapping section with young following restoration. Heatmaps are shown for the enriched sections of the top shared pathways (Figure 2D and Supplementary Figure 2J).

Figure 2. Molecular changes in age-related pathways following restoration. (A) Up- and down-regulated pathways in age, relative to young MPBs. (B) Enrichment plots for the top and down-regulated pathways. (C) Venn diagram shows shared and unique pathways between young and restored group, and overlap, changed pathways with restoration. (D) Enriched heatmaps of significant changes in ‘G’ (Full heat maps Supplementary Figure 2J). (E, F) Western blot for cell cycle proteins with extracts from 3 restored cells, MYC or E2F inhibitors or vehicle. (G) Clonogenic assay for CFU-GM with cells restored with 1 μM MDM2, 100 μM MYC or 40 μM E2F, mean ± SD (4 different aged donors, restored with 2 young donors; each in triplicates). See also Supplementary Figure 2). (H) Overview of NSG transplanted with aged huCD34+ MPBs. At the achievement of chimera, NSG with an aged hematopoietic system are injected with autologous restored (n = 12) or unrestored (n = 11) CD3-depleted MPBs. Serial transplantation used wk 33 huCD34+ cells. Controls were given young CD34+ cells (n = 8). (I–K) Flow cytometry for huCD45+ cells in BM and blood, ^*p < 0.05 vs. the other groups, (I), huCD34+ cells in BM (J), huCD3+ and huCD33+ cells in blood (K), mean % cells ± SD. (L) Lymphoid (CD3⁺+ CD19⁺)/myeloid (CD33⁺) ratio in BM and blood. (M) CFU-GM in cultures with huGM-CSF and huIL-3 and huCD45+ cells from BM, mean ± SD. (N) SASF array with huNSG plasma. Semi-quantitative densitometry used 1.5-fold cutoff for classification as up- or down-regulated, or no change. Heatmaps and piechart for differential gene expression. (O) RNA from huCD45+ BM cells evaluated qPCR gene arrays. Normalization to housekeeping genes used 1.5-fold cutoff, mean ± SEM. ^*p ≤ 0.05 vs. control. See also Supplementary Figures 2 and 3.

MYC, E2F and p53 in hematopoietic restoration

This section validated key pathways identified in the RNA-Seq analyses. MYC pathway, which was increased with restoration could explain hematopoietic restoration by partial reprogramming [29, 48]. E2F, due to its role in cell cycle regulation, which is important for hematopoietic homeostasis. We selected p53 because of increased MDM2 in the restored cells (Figure 1I, 1J and 1O). Reduced p53 would mitigate cellular senescence. Heterochronic cultures, in the presence or absence of MYC, E2F or MDM2 inhibitors or vehicle, indicated decreases in proteins linked to G1 transition by Western blot (Figure 2E and 2F). Clonogenic assay for CFU-GM in the cultures showed significant (p < 0.05) decreases with the inhibitors relative to vehicle (Figure 2G). In summary, decreased p53, and increases in MYC and E2F pathways are important for restoring the function of aged MPBs.

In vivo restoration of the aged hematopoietic system – humanized NSG (huNSG)

This section recapitulated the in vitro restoration in huNSG mice carrying aged hematopoietic system. We achieved hematopoietic chimera in NSG mice by transplanting aged CD34+ cells using dose-response and time-course studies. Mice began to achieve chimera after 8 weeks, based on human CD45+ cells in mice blood (Supplementary Figure 3A and 3B, depicted baseline chimerism prior to the second transplant). Mice with stable chimera (wk 19) were transplanted with CD3-depleted restored MPBs. Mice remained healthy until the end-point (Figure 2H, Supplementary Figure 3C–3H). After wk 14, we analyzed the blood, spleen and BM of mice for immune cells by flow cytometry (Supplementary Figure 3I–3N). Mice transplanted with restored cells showed significantly (p < 0.05) more huCD3 and huCD34+ cells, decreased myeloid cells and increased lymphoid:myeloid ratio, as compared to mice given unrestored aged cells (Figure 2I–2L). Clonogenic assays for CFU-GM using huCD45+ cells from mice femurs indicated a significant (p < 0.05) increase when mice were given restored cells as compared to those given unrestored aged cells (Figure 2M). These improvements correlated with enhancement in senescence profile: plasma, 12% restored vs. unrestored cells; qPCR senescence/aging array with femur cells, 8%, restored cells vs. unrestored cells (Figure 2N and 2O, Supplementary Figure 3O–3Q).

We assessed the functional competence of HSCs by selecting CD34+ cells from the experimental mice and transplanting into naïve NSG mice. CD34+ cells from mice with restored cells reconstituted NSG mice at week 8 whereas similar transplant with CD34+ cells from mice given aged cells did not achieve chimera (Figure 3A and 3B). In summary, transplantation of in vitro restored cells improved hematopoietic function in NSG mice carrying an autologous aged hematopoietic system.

Figure 3. Restored cells transplanted in NSG mice with aged human hematopoietic system (huNSG) and exosomal RNA in restoration. (A and B) HuCD34+ cells (10⁵), pooled from wk 33 mice (Figure 2H) were injected into naïve NSG mice (n = 3). At 12 wks, mice were analyzed for huCD45+ and huCD34+ cells. ND = none detected. (C) Pooled MVs (10⁶) from heterochronic or isochronic (young and aged) cultures were added to naïve aged MPBs on day 0 and 4 and at day 7, CFU-GMs were assessed in clonogenic assays, mean ± SD, n = 5. (D) Total RNA in MVs isolated from heterochronic cultures, ng^-8/MV ± SD, n = 8. (E) BCI-137 or vehicle was added to heterochronic or isochronic cultures, CFU-GM ± SD, n = 5. (F) 3D plots with data from qPCR miR array data using RNA from MVs. (G) Venn diagram showing differential and overlapping miRNAs. (H) qPCR for differentially expressed MV miRs. Shown are the consistently upregulated miR in young isochronic (dark green bar) and heterochronic cultures (striped bar), mean ± SD, n = 3. Aged isochronic cultures were assigned a value of 1. ^*p ≤ 0.05 vs. control. (I) MiRnome sequencing used small RNA from MVs of aged and young MPBs or UCB. Heatmaps used miRNA, > 1.4-fold between aged and young samples. Venn diagrams depicts the differential and overlapping miRNAs. (J, K) Studies, similar to `I’, compared miRNAs, sequenced from MVs of aged isochronic and heterochronic (cultured with young MPB or UCB) samples. (L) The 12 miRs showing differential expression between aged and young in ‘I’ were compared to miRs that were increased in heterochronic vs. aged isochronic cultures (I, J). Shown are the increased 8 miRs in restored cells. (M) qPCR for the 8 miRs, 7 biological replicates, each in triplicate. The data are normalized to miR-7641-2 and presented as fold change using 1 for aged control. (N) 6 validated miRs or control miRs were expressed, alone (left) or together (right) in 5 biological replicates, each in triplicate. CFU-GM at day 7, mean ± SD. ^*p ≤ 0.05 vs. control. See also Supplementary Figures 3 and 4.

Microvesicles (MVs) in aged hematopoietic restoration

Secreted MVs with their RNA, protein and lipid cargo could cross the culture membrane to establish communication between the aged and young cells [37, 49]. Also, MVs can regulate hematopoiesis [50]. We therefore determined if MVs, in particular exosomes, could be responsible for restoration of aged MPBs. We characterized the MVs from heterochronic and isochronic cultures and found similar size (Supplementary Figure 4A). Notably, aged MPBs produced higher number of MVs as compared to those released from heterochronic cultures or young MPBs (Supplementary Figure 4B). We also validated the incorporation of MVs into aged cells by added those from day 3 heterochronic cultures to naïve aged cells. The MVs, labeled with CMAC blue dye, were examined for transfer into the aged cells by 2D (EVOS imaging) and 3D (confocal microscopy) imaging (Supplementary Figure 4C; Supplementary Videos 1 and 2).

We tested the role of MVs in aged hematopoietic restoration by pooling those from day 4 and 7 heterochronic cultures. The pooled MVs (10⁶ particles, based on dose response studies) were added to naïve aged MPBs at seeding (time 0) and after 4 days. After 7 days, clonogenic assay for CFU-GMs indicated significantly (p < 0.05) more colonies with MVs from heterochronic cultures, relative to those from isochronic young and aged cultures, solidifying a role for MVs in hematopoietic restoration (Figure 3C).

We sought the candidate MV cargo by assessing the RNA content, which was higher in young MPBs as compared to aged MPBs (Figure 3D). We focused on miRNAs due to their role in hematopoiesis [51]. The pharmacological agent, BCI-137 (AGO2 inhibitor), blunted miRNA packaging during MV biogenesis without compromising their release (Supplementary Figure 4D, 4E). BCI-137 reduced the small RNA contents of MVs, including significant (p < 0.05) decrease of miRs, and decreased CFU-GM in the heterochronic cultures (Supplementary Figures 4F, 4G and 3E, 3F). These findings indicated a critical role for the miRNA content of MV in aged hematopoietic restoration.

We sought the causative miRNA(s) using an 84-probe array. MVs from isochronic (young and aged) and heterochronic co-cultures showed distinct miR profiles with 25 shared among the groups (Figure 3F and 3G). IPA predicted targets for the differentially expressed miRs and indicated a functional network in which young miRs are poised to prevent aging disorders such as cancer and hematological disorders (Supplementary Figure 4H). MiR-19a, -103a, -106b and -146a were upregulated in the young and restored MVs. Since qPCR only validated miR-19a in the young and restored cells (Figure 3H), we conducted further analyses by RNA-Seq of small RNA within the MVs. Using an expression cut-off of 100 mappable reads, we identified 13 and 17 unique miRs in aged and young MVs, respectively (Figure 3I). 12/17 miRs were higher in young vs. aged, while 3/17 were lower (Figure 3I, middle). The miRnome of MVs from UCB was vastly different with 70 miRs in >100 mapped reads but only 4 shared with young MPBs (Supplementary Figure 4I). Similarly, there were differences in intracellular miRnomes among aged, young, and UCB (Supplementary Figure 4J). Importantly, after restoration with young MPBs or UCB cells, there were common miR profiles in the heterochronic cultures, indicating that aged cells could be driving the MV cargo secreted from young cells (Figure 3J–3K, Supplementary Figure 4K–4O).

Among the 12 differentially expressed miRs between baseline young and aged MPBs, 8 were upregulated in the restored cells (Figure 3I and 3L). We validated the 12 miRs by qPCR, which were normalized with miR-7641-2 due to equal expression among the groups. These studies verified 6/12 miRs (Figure 3L, middle/M, Supplementary Figure 4M–4O). We then asked if the restored cells release MVs containing the validated miRs. To address this, we washed and then transferred the restored aged cells to the inner wells as facilitator young cells in heterochronic cultures (Supplementary Figure 4P). The MVs in these heterochronic cultures indicated the continued presence of miR-223 and -619 (Supplementary Figure 4P).

We performed cause-effect studies by expressing ≥1 of the 6 miRs in aged MPBs. These transfected cells were placed in cultures for 7 days, similar to the transwell cultures. Specifically, we cultured the aged cells to mimic restoration with young MPBs, except that these cultures dissected the restorative effects of specific miRNAs. After the culture period, we performed clonogenic assays for CFU-GM and noted significant (p < 0.05) increase in CFU-GM with miR-619 and/or -1303, and in combination with the other miRs (Figure 3N). Thus, miR-619, -1303 and -4497 became the restorative candidates for subsequent studies.

Downstream targets of restorative miRNAs

In order to understand how the MV-containing restorative miRs regulate hematopoietic restoration, we mapped the downstream effector pathways. First, we determined how miRs within the young MVs, and restored (heterochronic) or unrestored (isochronic) aged cells interact to cause functional changes. The miR interactome identified the top cellular pathways, including T-helper 1 and 2, cellular development and molecules that regulate senescence, e.g., CDKN2A and p53 (Figure 4A, Supplementary Figure 4Q and 4R). The greatest convergence between the two datasets was p53 (Figure 4B, orange lines). This outcome is in line with our studies showing p53 with a major role in hematopoietic restoration (Figures 1 and 2). We observed similar findings with the dataset from MVs released from heterochronic cultures using UCB (Supplementary Figure 4S–4V).

Figure 4. Exosomal miRNA targets in restoration. (A) IPA output of top predicted cellular functions (left) and canonical pathways (right) in analyses of MV miRNAs from the following: young vs. aged cells, heterochronic vs. aged isochronic cultures. (B) Radial depiction of young MV vs. restored intracellular interactome with p53 at the center of overlapping networks (Orange, direct interactions). (C) Analyses of 6 miRNAs (Figure 4L) for targets using TargetScan human database. (D) Targets were analyzed by IPA and the predicted networks (brown) compared to the young exosomal (blue) and aged heterochronic intracellular (dark orange) miRNA interactome. (E) Tabulation of selected targets and predicted interaction with the miRNA interactome. (F) 5 potential downstream targets for functional validation. (G, H) qPCR for candidate targets using RNA from aged cells of heterochronic or isochronic cultures (G), or human cells from femurs of huNSG (H), Fold change of normalized (β-actin) results, n = 4. (I) qPCR for PAX5 and PPM1F in aged and young MPBs, fold change with young donor assigned 1. (J, K) Aged MPBs were transfected with pre-miRs or control miR (J) and young MPBs, with anti-miRs or control miR (K). At day 7, the cells were analyzed for PAX5 and PPM1F mRNA by qPCR. The data are presented as the mean ± SD fold change, n = 4. The controls were assigned values of 1. (L) Aged MPBs were transfected with PAX5 or PPM1F siRNA or scramble (control). At day 7, the cells were analyzed for CFU-GM. Positive ctrl: heterochronic cultures, mean CFU-GM ± SD, n = 4. (M) Effect of PAX5 or PPM1F knockdown by siRNA on T-cell activation (CD25) for CD4⁺ (top panels) and CD8⁺ (bottom panels) populations. Right panels represent the % activated vs. total T-cells shown in orange. See also Supplementary Figure 4.

We applied miRNA target prediction software for direct targets of the 6 restorative miRNAs. Predicted targets (6101) for the individual miRNAs were stratified for common targets if they include the 6 miRs (Figure 4C). After this, we narrowed the targets if they share >3 miR hits and these hits must include the restorative miR-619 and -1303 (Figure 3). The remaining targets were scanned for known expression in relevant tissues. We eliminated targets encoding hypothetical proteins and if their expression was restricted to neural tissues. We then subjected the targets to predicted pathway analysis (Figure 4D, brown) in the context of our effector-target interactome (Figure 4D - blue and light orange, Figure 4E). The resulting 5 target candidates included the transcription/differentiation PAX5, which was increased in the unrestored isochronic cells, but significantly (p < 0.05) decrease after restoration (Figure 4F and 4G) [34]. Both PAX5 and phosphatase PPM1F, which is linked to integrin-mediated adhesion, were decreased in the femurs of restored mice (Figure 4H) [35]. These results led us to examine baseline expression of PAX5 and PPMIF. The results indicated consistent increase in PAX5 in the aged cells, but decrease in PPM1F, perhaps due to its complex function (Figure 4I).

We performed cause-effect studies between the restorative miRs and PAX5/PPM1F expression by transfecting the aged MPBs with pre-miR-619 or pre-1303 and -4497 (combo) or miR mimic (control). Analyses for PAX5 and PPM1F mRNA by qPCR indicated that miR-combo significantly (p < 0.05) decreased PAX5 and PPM1F mRNAs, compared to control or pre-miR-619 alone (Figure 4J). In corollary studies, we transfected young MPBs with anti-miR-619 or anti-miR (Combo) and observed increases in both PAX5 and PPM1F with miR-619 (Figure 4K).

Functional studies used aged MPBs, transfected with PAX5 or PPMIF siRNA or scramble siRNA followed by clonogenic assay for CFU-GM. PAX5 siRNA caused a significant (p < 0.05) decrease in CFU-GM whereas PPM1F siRNA significantly (p < 0.05) increased CFU-GM (Figure 4L). Positive control with heterochronic cultures resulted in the expected colonies. Knockdown of PAX5 and PPM1F led to T-cell activation, which is in line with improved lymphoid functions during restoration (Figure 4M).

In vivo hematopoietic restoration by candidate miRNAs

We asked if the restorative miRs (-619 and/or −1303 + −4497 (combo)) (Figure 3) could replace cellular in vitro restoration in heterochronic cultures. We addressed this by transplanting aged MPBs, transfected with the restorative miRs in NSG mice carrying aged hematopoietic system (Figure 5A). We selected mice with >1% blood chimera (huCD45+) to transplant aged MPBs that were transfected with miR-combo, miR-619 or control miR mimic (Supplementary Figure 5A and 5B). The mice showed normal pathology (Supplementary Figure 5C–5H). Compare to control miR, there was significantly (p > 0.05) more huCD45+ in the BM of mice with cells transfected with the restorative miRs (Figure 5B). Relative to control miR, miR-619 transfectants resulted in significantly (p < 0.05) more CD3+ and CD4+ cells and significant (p < 0.5) decrease of CD8+ cells (Figure 5C). The lymphoid:myeloid ratio, which is a hallmark of reverse aging was increased, with significant (p < 0.05) increase of CD19+ cells and decreased (p < 0.05) CD33+ myeloid cells (Figure 5D–5F).

Figure 5. Hematopoietic restoration with miRNAs. (A) Restoration protocol with NSG was similar to Figure 2H. Chimeric mice were given autologous CD3-depleted aged MPBs, transfected with miR-619, miR-combo, −619, −1303 and −4497, or control (RNA mimic) and then cultured for 7 days, (n = 18). (B–E) Mice were analyzed for huCD45 in BM and blood (B); huCD3, CD4 and CD8 in blood (C); HuCD19 in blood and BM (D); HuCD33 in BM (E). Results presented as % mean cells ± SD (F) Lymphoid:Myeloid ratio (CD3⁺+CD19⁺/CD33⁺) in BM. (G) CFU-GM with huCD45+ cells from femurs, mean ± SD. (H) qPCR for PAX5 and PPM1F with RNA from huCD45+cells from femurs. Fold change ± SD used 1 for the lowest value. (I, J) RNA from `H’ were analyzed in qPCR human senescence and aging arrays. Normalized results used 1.5-fold cutoff to classify up- or down-regulation, or no change (I). SASF analyses with plasma. Semi-quantitative densitometry used 1.5-fold cutoff, similar to I (J). Differential gene and protein expressions as heatmaps (left), pie charts (top) and bar graph (bottom), mean ± SD. ^*p ≤ 0.05 vs. control. See also Supplementary Figure 5.

Clonogenic assay for CFU-GM using huCD45+ cells from mice femur indicated significant (p < 0.05) increase when mice were given miR-combo- or miR-619-transfected aged cells as compared to the group given control miR mimic (Figure 5G). PAX5 level was significantly (p < 0.05) decrease in mice with miR-619 or miR-combo transfectants (Figure 5H). Similar decrease was observed for PPM1F, but only for miR-combo (Figure 5H).

The miR-mediated restoration also reduced cellular senescence, based on PCR array with cDNA from huCD45+ femur cells (Figure 5I, Supplementary Figure 5I–5K). MiR-combo significantly (p < 0.05) decreased (>50%) the senescence/age-related genes. SASP factors in mice plasma showed 51.4% and 45.6% decreases for miR-619 and miR-combo, respectively (Figure 5J). Taken together, the miRs restored hematopoiesis similar to cells from heterochronic co-cultures, and highlighted roles for the transcription factors PAX5 and PPM1F in restoration.

Enhanced Natural killer (NK) activity within restored old MPB in immune therapy

Age-related defects in NK cells can compromise surveillance for emerging malignancy, infection and induce cell senescence [36, 52]. We asked if the restoration process corrected the age-linked NK defects. Flow cytometry indicated 8-fold more CD56+ cells within the restored cells, relative to unrestored cells (Figure 6A). This correlated with significant (p < 0.05) increase in Lytic Units (Supplementary Figure 6A) in 3/4 aged restored cells (Figure 6B).

Figure 6. Restorative containing NK cell function. (A) Flow cytometry for CD56+ cells, pre- and post-restoration. (B) Timeline LUs (Supplementary Figure 6A for calculation) in restored cells. (C) Treatment protocol with mice harboring dormant CSCs. (D) Flow cytometry for huCD45+ cells in blood of mice after 1 month of injection with CD3-depleted restored cells. (E) qPCR for huGAPDH at 2 months after injection with CD3-depleted restored cells, mean ± SD (n = 5); ND = not detected. (F) CFU-GM and BFU-E in BM at 1 month post-restoration, mean ± SD, n = 5. ^*p ≤ 0.05 vs. mice with unrestored cells. (G) Survival curve spanning the study period. (H) At yr 1, qPCR for GFP with cells from femurs. The results presented as fold change in which the lowest value was assigned 1, mean ± SD, 4/group. (I) CSCs, co-cultured with restored MPBs (− or + NK cells) for 24 h. % CSCs (GFP+) were determined by microscopy and flow cytometry, mean ± SD, n = 4. (J) Protocol for NSG with dormant CSCs given restored CD3(−) MPB (−/+ NK cells). (K–M) GFP (surrogate of CSCs) in paraffin section of femurs, −/+ restored MPB: -MPB (K), +MPB (L), MPB without NK cells (M). (N) Summary: Aged hematopoietic cells instructed young cells to produce specific miRNA containing MVs to induce MYC and E2F targets to restore the aged MPBs, leading to increased NK activity. Transplantation of restored cells decreased hallmarks of aging: ↑lymphoid:myeloid, ↓senescence/inflammation. See also Supplementary Figure 6.

We further examined the function of the restored NK cells. We selected a model of dormant breast cancer (BC) cells (BCCs) in nude BALB/c [53] (Figure 6C). At 1 month, the mice with CD3-depleted restored MPBs were chimeric for human hematopoietic function with no evidence of lethargy, no palpable lymph node and comparable survival between groups (Figure 6D–6G). Since the BCCs were transfected with pOct4a-GFP, we evaluated the mice femurs for GFP by qPCR. The results indicated 6 folds more GFP+ cells in mice given unrestored MPBs, related to mice with restored aged MPBs (Figure 6H). We did not detect GFP+ cells in the blood, indicating that the dormant cells remained in the femur (Supplementary Figure 6B and 6C). Together, these findings showed the restored MPBs capable of clearing dormant BCCs in the BM.

Since the dormant BCCs were cleared in nude mice, we performed cause-effect studies to determine if restoration of NK functions could explain the dramatic decrease in BCCs. First, we performed in vitro studies by co-culturing BC stem cells (CSCs/GFP+) with restored aged MPBs, with or without NK cells (Supplementary Figure 6D). The number of GFP+/CSCs was significantly (p < 0.05) reduced with NK-containing restored MPBs as compared to NK-depletion (Figure 6I). Adapting these studies in NSG mice (Figure 6J), we could not detect GFP+ cells in mice given restored NK+ cells whereas control mice (baseline) and mice given NK- restored cells showed dormant CSCs (green) in femurs (Figure 6K–6N). In summary, NK cells in restored MPBs effectively targeted and mostly cleared dormant BCCs in vivo.

Cell lines

K562 and MDA-MB-231 were cultured as per American Type Culture Collection instruction. The cell lines were tested by Genetica DNA Laboratories (Burlington, NC) and were deemed to be the original cells using ATCC STR database (https://www.atcc.org/search-str-database).

Human subjects

The use of de-identified human mobilized peripheral blood (MPB), peripheral blood (PB) and umbilical cord blood (UCB) was approved by Rutgers Institutional Review Board (IRB), Newark, NJ, USA.

Isolation of CD34⁺ cells and cell cryopreservation

Total MPB, herein referenced as total nucleated cells (TNC), were cryopreserved by adding chilled cryopreservation media to the cell suspension at a 1:1 ratio while gently shaking the cells. The resuspended TNC was ~ 5 × 10⁷ cells/mL. Cells were stored using a controlled rate freezer (Cryo Met Freezer, Thermo Fisher) at −1°C/min until the temperature reached -100°C. After this, the cells were transferred into liquid nitrogen.

CD34⁺ cells were selected using the CD34 microbead human kit (Miltenyi, Auburn, CA, USA). The method followed manufacturer’s instructions. Briefly, TNC was pelleted at 4°C, 10 min, 500 g. The TNCs were resuspended in cold MACS buffer at a concentration of 10⁸ cells/300 μL buffer. Cell suspension was incubated in 100 μL of FcR Blocking Reagent and CD34 MicroBeads for 30 min at 4°C. After this, cells were washed with MACS buffer and the CD34⁺ cells were positively selected by magnetic separation with LS columns. The purified CD34⁺ cells were immediately cryopreserved as described above.

Transwell assay

Non-contact cultures were performed in 6- or 12-well transwell system, separated by a 0.4 μ membrane (BD Falcon, ThermoFisher). In the case of 6-well plates, the inner wells contained 10⁷ young MPB or CB with the same number of aged MPBs in the outer wells. The numbers were scaled lower for 12-well plates. At optimization, cells were cultured for up to 5 wks in RPMI 1640 supplemented with 10% FCS, 2 mM glutamine and 0.5 μM β-mercaptoethanol (β-ME) (R10 media). Cultures with allogeneic cells were designated heterochronic and those with autologous cells, isochronic. During the culture period, 50% percent of cell-free media were replaced every four days with fresh R10 media without cell passaging. Except for FCS, we did not supplement the media until we were ready to evaluate aged hematopoietic activity in clonogenic assay. The latter contained supplement (see below for method). In this case, we did not mask the restoration with exogenous growth factors. Based on the optimization data, all other transwell cultures were conducted for 1 wk.

Cell migration from inner to outer wells

Transfer of cells from the inner to the outer wells of co-cultures were studied as follows: young MPBs were labeled with CarboxyFluorescein succinimidyl ester (CFSE) (Thermo Fisher) following manufacturer’s recommended protocol. Briefly, CFSE was diluted in 1x PBS to 5 μM (stock solution) and then diluted at 1/1000 (working solution). The latter was used to label cells by incubating for 20 min at room temperature. The labeled cells were added to the inner wells of the heterochronic cultures and at weekly intervals, the cells in the outer wells were analyzed for CFSE by flow cytometry.

Flow cytometry

Cells (10⁶) were labeled with the primary antibodies in 100 μl PBS. Each labeling used an isotype IgGs for background labeling (control) at the same concentration as the primary antibodies (see above for concentrations). After 30 mins at room temperature, the cells were washed with PBS and acquired on a FACSCalibur flow cytometer (BD Biosciences). Data were analyzed using BD CellQuest Pro™ software (BD Biosciences).

Western blot analyses

Cell extracts were isolated with cell lysis buffer (50 mM Tris-HCl (pH 7.4), 100 mM NaCl, 2 mM MgCl₂, 10% glycerol, and 1% NP-40) as described (Ghazaryan et al., 2014). Extracts (15 μg protein) were electrophoresed on SDS-PAGE and then transferred to polyvinylidene difluoride membranes (Perkin Elmer). The membranes were incubated overnight at 4°C on rocker with primary antibodies in 5% milk dissolved in 1x PBS tween. Next day, the membranes were incubated with species-specific HRP tagged secondary IgG in the same diluent. After 2 h, the membranes were developed by chemiluminescence using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific).

Clonogenic assays

Clonogenic assays for hematopoietic progenitors, granulocytic-monocytic (CFU-GM) and early erythroid (BFU-E), were performed as described in methylcellulose matrices (Rameshwar et al., 2001). Cultures for CFU-GM contained 3 U/mL GM-CSF and BFU-E, 3 U/mL rhIL-3 and 2 U/mL Epo. Colonies were counted by a blinded observer. Each colony contained >15 cells.

Modified long-term culture-initiating cell (LTC-IC) assay

Confluent stromal cells in the outer well of the transwell cultures were subjected to 1.5 Gy, delivered by a cesium source. After 16 h, the floating cells were washed and 10⁷ BM mononuclear cells added to the γ-irradiated stroma. At wks 6, 10 and 12, aliquots of mononuclear cells were assayed for CFU-GM in clonogenic assays (see above) [61].

Mixed lymphocyte reaction (MLR)

One-way mixed lymphocyte reaction (MLR) was performed as described previously [62]. Briefly, γ-irradiated (20 Gy) stimulator cells from isochronic or heterochronic cultures were added to 96 well-flat bottom plates (Corning, Corning, NY) at 10⁶/mL in 0.1 mL volume in triplicates. Autologous responder freshly thawed cells were added at the same concentration and volume. Control MLR used peripheral blood mononuclear cells (PBMCs) from two different donors (20–30 yrs). The MLR reactions were incubated at 37^oC. At 72 h, the wells were pulsed with 1 μCi/well of [methyl-³H]TdR (70–90 Ci/mmol; NEN Radiochemicals, PerkinElmer, Akron, OH). After 16 h, the cells were harvested with a PhD cell harvester (Cambridge Technologies, Bedford, MA) onto glass-fiber filters, and [³H]TdR incorporation quantified in a scintillation counter (Beckman Coulter, Brea, CA, USA). The results are expressed as the stimulation index (S.I.), which is the mean dpm of experimental cultures (responders + gamma-irradiated stimulators)/dpm of responder cells cultured in media alone.

Cell titer blue viability assay

The cells in the outer wells of transwell cultures were assayed for viability with CellTiter-Blue Cell Viability Assay kit (Promega, Madison, WI). Viability analyses followed manufacturer's instructions. Briefly, CellTiter-Blue reagent was added to aliquots of cells in 96-well plates. The plates were incubated for 4 h at 37°C and then read on a fluorescence microplate reader at 560 nm excitation/590 nm emission. Percent viability was calculated from a reference using untreated healthy cells, which were considered 100% viable, and cell-free wells containing reagent alone, which were considered 0% viable.

Real time PCR for human cells in mouse femur

RNA extraction was performed according to manufacturer’s protocols with the RNeasy Mini Kit (Qiagen, Germantown, MD). Quality and concentration of RNA were determined with the Nanodrop ND-1000 spectrophotometer (ThermoFisher Scientific). The High-Capacity cDNA Reverse Transcription Kit (Life Technologies) was used to convert RNA to cDNA and 10 ng used in real-time PCR with Sybr Green PCR Master Mix II (Life Technologies) using primers from human and mouse GAPDH (see above).

PCR array

Total RNA (2 μg) was extracted from MPB with the RNeasy Mini Kit (Qiagen) and then reverse-transcribed using the First Strand cDNA Synthesis Kit (Life Technologies). The cDNA (200 ng) was used for quantitative PCR with the Human Cellular Senescence RT² Profiler™ PCR Array (Qiagen) containing 84 key genes involved in the initiation and progression of the biological process causing cells to lose the ability to divide. Prepared cDNA templates were added to the ready-to-use PCR master mix at equal volumes within each well of the same plate. The arrays were then run on the 7500 Real Time PCR System (Life Technologies) with the cycling profile (40 cycles) as follows: 94°C for 15 secs and 60°C for 45 secs. The data were analyzed with Qiagen PCR Array Data Analysis Software and normalized to five housekeeping genes provided within each array. Controls were also included for genomic DNA contamination, RNA quality, and general PCR performance. The normalized data are presented as fold difference, with a value of 1 representing no change. Differential and overlapping miRNAs are presented in Venn diagram. The overlapping areas include miRNAs with <1.5-fold difference among groups.

MV isolation and nanoparticle tracking analysis

MVs were isolated from cell culture media as described (Bliss et al., 2016). In addition, MVs were also isolated with the Total Exosome Isolation Kit (Life Technologies), using a modified version of the manufacturer’s protocol. Specifically, isochronic and heterochronic cultures were established with Exosome-depleted FBS Media Supplement (System Biosciences, Palo Alto, CA, USA). Culture media collected on the 4th and 7th days were pelleted and supernatant transferred to another tube for further clarification at 2000 g for 30 min to remove residual cells and debris. The remaining supernatant was transferred to a fresh tube and 0.5 volumes of Total MV Isolation reagent was added for overnight incubation at 4°C. The following day, samples were centrifuged at 10,000 g for 1 h at 4°C to pellet the exosomes for subsequent nanoparticle tracking analysis (NTA) or long-term storage at −80°C. Analysis of absolute size distribution of MVs was performed using the NanoSight LM10 with NTA3.1 software (Malvern Panalytical, Malvern, UK). Particles were automatically tracked and sized based on Brownian motion and the diffusion coefficient. For NTA, MVs were re-suspended in 0.5 mL of PBS and measured using the following parameters: Temperature = 25.6 +/− 0.5°C; Viscosity = (Water) 0.99 +/− 0.01 cP; Measurement time = 30 sec; Syringe pump speed = 30. The detection threshold was similar in all samples. Three recordings were performed for each sample.

MV entry into aged cells

Purified MVs (4 × 10⁶/500 μL) from 3-day co-cultures were resuspended in PBS containing 25 μM CellTracker Blue CMAC Dye (ThermoFisher Scientific). The MVs were incubated for 45 min at 37°C. After this, MVs were washed as follows: Increased PBS to 5 mL followed by ultracentrifugation at 130,000 xg for 80 mins in the cold. The MVs were then added to naïve aged MPBs and then subject to timeline (24 and 48 h) for cellular uptake. The latter was assessed by the following methods: Confocal 3D imaging using Olympus Fluoview FV10i and the images processed with the FV10-ASW software, Version 04.02.03.06; 2D imaging with the EVOS FL Auto2 (Invitrogen/ThermoFisher Scientific).

MitoSox assay

MPB cells (10⁶) were labeled with anti-CD34-APC and -CD45-PerCp-Cy5.5, as described below for flow cytometry. The cells were washed and incubated with 5 μM MitoSox™ Red for 10 min at 37°C in the dark and then washed again with warm HBSS. The data was acquired and analyzed with the FACSCalibur.

Senescence Associated Secretory Phenotype (SASP) array

Senescence and aging gene arrays

Cells were flushed from the femurs of huNSG using a 26-gauge needle. Murine cells were eliminated with a Mouse Cell Depletion Kit. Total RNA (2 μg) was extracted from purified cells using the RNeasy Mini Kit and reverse-transcribed with the RT² First Strand Kit. 20 ng of cDNA was used for qPCR with the Human Cellular Senescence and Human Aging RT² Profiler™ PCR Arrays (Qiagen). Arrays were run on the 7300 Real Time PCR System (Life Technologies) with the cycling profile (40 cycles): 94°C for 15 secs and 60°C for 45 secs. Gene expression analysis was performed using Qiagen PCR Array Data Analysis Software. The data were normalized to five housekeeping genes provided within each array. The data was submitted to the NCBI Genome Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) under SuperSeries accession number GSE138565. Hierarchical clustering and heat map generation were performed with Heatmapper software (http://www.heatmapper.ca/), as described [64].

MiRNA microarray and qPCR

Total RNA (500 ng) was isolated from MVs using the miRCURY RNA Isolation Kit (Exiqon). RNA was reverse-transcribed with the miScript II RT Kit (Qiagen) and 20 ng of cDNA used for qPCR with the human miFinder miRNA Array (Qiagen). The PCR was done with the following cycling conditions: 94°C for 15 mins, 40 cycles at 94°C for 10 secs, 55°C for 30 secs, 70°C for 30 secs, followed by melt curve analysis. The data were analyzed with the online miScript miRNA PCR Array data analysis tool (Qiagen).

Validation of miRNAs identified in the microarray and NGS was done by individual qPCR experiments using miScript primer assays and similar cycling and analysis schemes. Total RNA (2 μg) was also isolated from cells, as described above, for profiling of downstream miRNA targets by qPCR using the primer pairs listed above with the following cycling conditions: 95°C for 15 mins, 40 cycles at 94°C for 15 secs, 51°C for 30 secs, 72°C for 30 secs, followed by melt curve analysis. Analyses were performed with Qiagen PCR Array Data Analysis Software, as described above. Array and individual qPCR studies were normalized to RNU6, SNORD68 and SNORD95 and presented as fold change. The 30 genes with marked changes following restoration were selected for links to other genes using RNA-Protein Association and Interaction Networks (RAIN) [65].

RNA-Seq

RNA Seq was performed at the Genomics Center at Rutgers New Jersey Medical School (Newark, NJ, USA). Total RNA was submitted to the center where poly A RNA was purified using NEBNext^® Poly(A) mRNA Magnetic Isolation Module (New England BioLabs, Ipswich, MA, USA). Next generation sequencing (NGS) libraries were prepared using the NEB Ultra II Library Preparation Kit and NEBNext^® Multiplex Oligos for Illumina (Dual Index Primers Set 1) (New England BioLabs). The generation of the libraries followed manufacturer’s protocol. The libraries were subjected to quality control using Qubit instrument and high sensitivity Kit from Thermo Fisher as well as Tapestation 2200 instrument and D1000 ScreenTapes from Agilent (Santa Clara, CA, USA). The libraries were diluted to 2 nM and then denatured as per Illumina’s protocol and run on Illumina’s NextSeq instrument (San Diego, CA, USA) using 1X75 cycle high throughput kit. The BCL files that were generated from the sequencing were demultiplexed and converted to FastQ files using BCL2FASTQ software from Illumina. All raw and processed sequencing data have been submitted to the NCBI Genome Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE138563, SuperSeries accession number GSE138565.

MiRNA sequencing (miR Seq)

Total RNA from MVs and cells was isolated using the miRCURY RNA Isolation Kit with small and large RNAs fractionated with the RNeasy MinElute Cleanup Kit, both according to manufacturer’s recommended specifications (Qiagen). Half of the small RNA fraction (200 ng) was used in library preparation with the NEBNext Multiplex Small RNA Sample Prep Set for Illumina - Set 1 (New England Biotechnology), according to the following protocol: (1) ligation of the 3′ SR Adaptor, (2) hybridization of the reverse transcription primer, (3) ligation of the 5′ SR Adaptor, (4) reverse transcription for first strand cDNA synthesis and (5) PCR enrichment. After PCR, samples were cleaned up and size selection performed. Briefly, 2 μl of sample was subjected to TapeStation (Agilent) analysis to ascertain band sizes. Samples were run on 8% acrylamide gel at 100V for 1 hr, with correct size bands excised for gel purification. Small RNA libraries were diluted to 2 nM and run on a miSeq System (Illumina) for NGS using the V2 kit (Illumina). All raw and processed sequencing data have been submitted to the NCBI Genome Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE138564, SuperSeries accession number GSE138565.

Data analyses

Nucleofection of miRNA mimics, miRNA inhibitors and siRNA

Aged cells (10⁷/sample) were nucleofected with miRNA mimics (Qiagen), miRNA inhibitors (Qiagen), negative control siRNA (Qiagen), negative control miRNA inhibitor (Qiagen) or downstream target candidate siRNAs (Origene) using the Amaxa P3 Primary Cell 4D-Nucleofector X Kit (Lonza) on a 4D Nucleofector device (Lonza), according to manufacturer’s specific protocol. Briefly, CD34⁺ cells were nucleofected with 60 nM of total miRNA mimics, miRNA inhibitors or siRNA using the “human CD34⁺ cell” program. Real time PCR for miRNA levels were similar (p > 0.05) between restored CD34+ cells and those subjected to nucleofection.

Sorting of breast cancer stem cells (CSCs)

MDA-MB-231 breast cancer cells were stably transfected with pEGFP1-Oct3/4, as previously described [53]. The CSCs were selected as the top 5% cells with the highest GFP intensity (Oct4^hi) and then sorted with the FACSAria II cell sorter (BD Biosciences).

In vivo studies

The use of mice was approved by Rutgers Institutional Animal Care and Use Committee (Newark Campus, NJ). Mice were housed in an AAALAC-accredited facility.

Immunohistochemistry

Immunohistochemistry for human pan cytokeratin and Ki67 in murine femurs was performed as described [37]. Briefly antigen was retrieved from deparaffinized sections of mouse femurs by heating at 56°C overnight. After this the slides were dewaxed with xylene and ethanol then rehydrated, fixed and permeabilized with 0.1% Triton X. Slides were then washed 3x in 1x PBS and then incubated overnight at 37°C in a humidified chamber with primary antibodies at 1/500 final dilution for anti-cytokeratin and anti-Ki67. The slides were washed and then incubated (2 hrs at room temp) in a humidified chamber with fluorescence-tagged secondary antibodies at 1/1000 final dilution: AlexFluor 405 for cytokeratin and rhodamine for Ki67. The slides were washed then covered with 1x PBS and then immediately analyzed on the EVOS FL Auto 2 Imaging System (ThermoFisher Scientific). Tissues from scraped femurs were placed on slides and then similarly imaged.

Co-culture between natural killer (NK−/+) and CSCs

NK cell depletion in the restored and unrestored MPBs was conducted with human CD56 microbeads from Millitenyi (Auburn, CA, USA). The method followed manufacturer’s instruction. Co-cultures were performed with 10⁵ breast CSCs and 10⁷ restored MPB, with or without NK cells. After 24 h, the number of CSCs were determined by the number of GFP+ breast cancer cells using fluorescence microscopy (Evos FL2 auto).

NK cell assay

T-Cell activation assay

T-cell activation was determined using the T-Cell Activation/Expansion Kit. Briefly, anti-biotin MACSiBead Particles were loaded with CD2, CD3, and CD28 antibodies. Cells from 7-day isochronic or heterochronic cultures, or MNCs isolated from huNSG mouse blood by Ficoll-Hypaque density gradient, were incubated with loaded anti-biotin MACSiBead Particles at a 1:2 bead to cell ratio for 72 h to activate T-cells. Addition of unloaded MACSiBead Particles served as negative control. After 72 h, cells were fluorescently labeled using CD45-FITC, CD4-PE, CD25-PerCP-Cy5.5 and CD8-APC to determine T-cell activation status by flow cytometry.

Statistical analyses

Statistical analyses were performed with ANOVA and Tukey-Kramer multiple comparisons test. For array and NGS expression analyses, average linkage was used for clustering and Pearson correlation analysis used for distance measurement to generate heatmaps and hierarchically cluster genes. P < 0.05 was considered significant.