Research ArticleTransplantation

Enhanced human hematopoietic stem and progenitor cell engraftment by blocking donor T cell–mediated TNFα signaling

See allHide authors and affiliations

Science Translational Medicine  20 Dec 2017:
Vol. 9, Issue 421, eaag3214
DOI: 10.1126/scitranslmed.aag3214

TNFα tampers with stem cell success

Most stem cell transplantation procedures are performed with unrelated donor/recipient pairs. One source of stem cells is umbilical cord blood, but the number of cells derived from this source can be limiting. Wang et al. examined factors that affect proliferation, engraftment, and differentiation of human umbilical cord stem cells in a preclinical model. They found that donor T cell production of TNFα was harmful to stem cell health. In the future, inhibiting TNFα after stem cell transplantation could lead to improved patient outcomes.

Abstract

Allogeneic hematopoietic stem cell transplantation (HSCT) is a curative therapy, but the large number of HSCs required limits its widespread use. Host conditioning and donor cell composition are known to affect HSCT outcomes. However, the specific role that the posttransplantation signaling environment plays in donor HSC fate is poorly understood. To mimic clinical HSCT, we injected human umbilical cord blood (UCB) cells at different doses and compositions into immunodeficient NOD/SCID/IL-2Rgc-null (NSG) mice. Surprisingly, higher UCB cell doses inversely correlated with stem and progenitor cell engraftment. This observation was attributable to increased donor cell–derived inflammatory signals. Donor T cell–derived tumor necrosis factor–α (TNFα) was specifically found to directly impair the survival and division of transplanted HSCs and progenitor cells. Neutralizing donor T cell–derived TNFα in vivo increased short-term stem and progenitor cell engraftment, accelerated hematopoietic recovery, and altered donor immune cell compositions. This direct effect of TNFα on transplanted cells could be decoupled from the indirect effect of alleviating graft-versus-host disease (GVHD) by interleukin-6 (IL-6) blockade. Our study demonstrates that donor immune cell–derived inflammatory signals directly influence HSC fate, and provides new clinically relevant strategies to improve engraftment efficiency during HSCT.

INTRODUCTION

Allogeneic hematopoietic stem cell transplantation (HSCT) is a potentially curative therapy for patients with hematologic diseases. However, 70% of patients do not have a matched related donor (1). Finding matched unrelated donors is clinically beneficial but has proven challenging especially for the minority of patients (2). Umbilical cord blood (UCB) provides a source of transplantable HSCs with several advantages (3). However, the limited number of hematopoietic stem and progenitor cells (HSPCs) available in single UCB units is associated with delayed hematopoietic recovery and a higher risk of graft failure, which are major complications that contribute to transplant-related mortality (TRM) in adult patients during the early posttransplantation phase (4). Potential strategies to improve engraftment include transplantation of double UCB units (5, 6), ex vivo expansion of HSPCs (711), and increasing HSPC homing (12, 13). Despite encouraging results reported in early-phase clinical studies (9, 10, 1215), the requirement for high cell numbers to overcome apparent low stem cell engraftment efficiencies remains poorly understood.

Accumulating evidence demonstrates that transplantation of HSCs alone is not sufficient to ensure robust engraftment and patient survival (16, 17). Coinfusion of lineage-committed progenitors and mature hematopoietic and immune cells has been associated with accelerated hematopoietic reconstitution (10), prevention of opportunistic infections (18), and exerting graft-versus-leukemia effects (19). In addition, pretransplant conditioning appears to induce host tissue damage and inflammation, leading to donor immune cell activation and secretion of inflammatory factors and graft-versus-host disease (GVHD) (20). One promising therapeutic strategy to attenuate acute GVHD focuses on characterizing the secreted factor dynamics during the early posttransplantation phase and using specific factor-targeting biologics (21, 22). We previously demonstrated that cell feedback networks regulate HSC fate during homeostasis or in vitro culture (8, 2325). Here, we report that upon transplantation, donor immune cell–mediated inflammatory signals feedback to donor stem and progenitor cells and affect short-term HSC engraftment and hematopoietic recovery. Furthermore, specifically blocking these inflammatory signals using a monoclonal antibody enhances bone marrow (BM) reconstitution levels and diversity and accelerates peripheral hematopoietic recovery rates.

RESULTS

UCB cell dose and composition affect short-term progenitor cell engraftment

Purified human CD34+ cells have typically been used to model HSCT in preclinical mouse models. To better understand how cotransplantation of CD34+ and differentiated/immune cells affects donor HSPC engraftment, we transplanted single, unfractionated human UCB units into sublethally irradiated NOD/SCID/IL-2Rgc-null (NSG) mice (Fig. 1A). First, we tested UCB doses of 0.5 × 106, 1 × 106, or 5 × 106 cells to examine cell dose–dependent effects on hematopoietic reconstitution. Infusion of 2.5 × 104 CD34+ cells (>98% purity) isolated from the same UCB unit served as a control (tables S1 and S2) (26). This experiment was limited to 28 days because, after 17 days, mice transplanted with 5 × 106 UCB cells experienced significant weight loss (P = 0.034) and other GVHD-related symptoms according to the standard GVHD scoring metric (fig. S1) (27).

Fig. 1 Increasing UCB cell dose impairs short-term progenitor cell engraftment.

(A) Experimental scheme. (B) Fold changes in CD34+ and CD34+CD45RACD90+ hematopoietic stem cell–enriched (HSC-e) cells [calculated as cell number harvested from bone marrow (BM) of two femurs and two tibiae in relation to the number injected on day 0; see table S1 for the cellular composition of the purified CD34+ fraction and unfractionated umbilical cord blood (UCB) cells and table S2 for the cell number transplanted on day 0]. Dots represent individual mice, with horizontal lines denoting the mean values (n = 5 mice per group from one experiment). Two-tailed Kruskal-Wallis and Mann-Whitney U tests with Bonferroni correction (Padj = 0.992) were used to determine the significance level. *P < 0.05; **P < 0.01; and ***P < 0.001. (C) Heat map representation of the percentage of human cell subsets (gated on 7-AADCD45+HLA-ABC+) detected in the mouse BM (see table S3 for phenotypic definition and table S4 for P values).

We observed human (CD45+HLA-ABC+) cell engraftment in the BM by day 28 in all transplantation conditions (fig. S2A). Fold changes in the number of CD34+ and CD34+CD45RACD90+ HSC-enriched cells (hereafter referred to as HSC-e) in relation to the number transplanted decreased with increasing UCB doses (Fig. 1B). These changes correlated with greater proportions of donor-derived mature immune cells [T and natural killer (NK)] and lower proportions of progenitors [precursor NKT (preNKT), progenitor B (proB), and megakaryocytic progenitor (MegaP)] and mature cells of other lineages [B, megakaryocyte (Mega), granulocyte (Gran), and monocyte (Mono)] present in the BM (Fig. 1C and fig. S2). The inverse relationship between the input UCB cell dose and the engraftment level and the predominant presence of donor immune cells at higher UCB doses imply that both the UCB dose and the coinfused immune cells could affect the short-term engraftment level of phenotypically enriched stem and progenitor cells. These observations motivated us to investigate the effect of immune cells on donor HSPC fate after transplantation.

Secreted donor-derived inflammatory factors affect in vitro HSC-e cell proliferation and differentiation

We hypothesized that inflammatory factor secretion from donor immune cells, upon exposure to irradiation-induced damage in host tissue and alloantigens, negatively affects transplanted HSPC function. To test this, we first measured human factor concentrations in the sera of UCB-infused mice. An array of human inflammatory factors was elevated within 2 weeks of UCB (5 × 106) infusion (Fig. 2A and fig. S3). Notably, these factors were not detected when purified CD34+ cells were transplanted alone. Next, using our previously published bioinformatics strategy (25), we found that seven of the elevated inflammatory factors, whose levels were 10 times the assay detection limit (arbitrarily set threshold; >32 pg/ml), have cognate receptors expressed by human HSCs at the transcriptional level (Fig. 2A and table S5). Among these seven factors, interleukin-6 (IL-6), granulocyte-monocyte colony-stimulating factor (GM-CSF), vascular endothelial growth factor (VEGF), and tumor necrosis factor–α (TNFα) were examined in our previous studies using an enriched HSC population (LinCD34+CD38CD45RACD49f+) (25, 26). It was found that IL-6 (26) and GM-CSF (25) stimulate the proliferation and differentiation of these cells, that VEGF has no direct effect (25), and that TNFα inhibits HSC proliferation (25). Therefore, we sought to examine TNFα plus the three remaining factors—interferon-inducible cytokine (IP-10), interferon-γ (IFNγ), and monocyte chemotactic protein 1 (MCP-1)—at concentrations relevant to those detected in the mouse serum using the same assay described previously (Fig. 2B) (25). No effects were detected for IP-10 and MCP-1 compared to the “3F” condition containing stem cell factor (SCF), thrombopoietin (TPO), and FMS-like tyrosine kinase 3 ligand (Flt3L) (hereafter referred to as control) (Fig. 2C and fig. S4). In the presence of IFNγ (10 ng/ml), the number of HSC-e decreased, whereas the number of differentiated cells increased, suggesting that IFNγ acts in a differentiation-inducing manner. The presence of TNFα resulted in a significant reduction of both HSC-e and progenitor cell compartments at 1 ng/ml (P < 0.01). It is worth noting that the differentiated cell population was only affected by TNFα at concentrations greater than 1 ng/ml, as we previously showed (25). These data indicate a dose-dependent effect of TNFα on specific cell populations at different developmental stages.

Fig. 2 Systemically released inflammatory factors directly affect HSC-e cell fate.

(A) Concentrations of human factors detected in the mouse serum after 14 days. Seven factors whose corresponding receptors were differentially expressed by HSCs are indicated as bold. (B) Forty LinCD34+CD38CD45RACD49f+ cells were cultured in the serum-free control condition with the addition of each test factor. After 7 days, changes in the numbers of CD34+CD45RACD90+ (defined as HSC-e) and the remaining CD34+ (progenitor) and CD34 (differentiated) cells were calculated for each test condition against control. APC, allophycocyanin; FITC, fluorescein isothiocyanate; PE, phycoerythrin. (C) Average values of fold change in relation to control when interferon-inducible cytokine (IP-10) (orange), monocyte chemotactic protein 1 (MCP-1) (green), interferon-γ (IFNγ) (blue), and tumor necrosis factor–α (TNFα) (purple) were individually tested at three different doses (ng/ml) indicated by different shades for each factor. Two-tailed Kruskal-Wallis test and Mann-Whitney U test with Bonferroni correction (Padj = 0.983) were performed to compare each test condition to the control group. Data are from four or five independent experiments (n = 3 within each experiment). Two to five pooled UCB units were used per experiment. *P < 0.05; **P < 0.01; ***P < 0.001; and not significant (n.s.), P ≥ 0.05.

TNFα inhibits HSPC proliferation via TNF receptor 1

To directly study the effects of TNFα on HSPCs, purified CD34+ cells were cultured with TNFα (1 ng/ml) for a 7-day period. We observed that both the percentage and fold expansion of CD34+ and CD34+CD45RACD90+ HSC-e cells were significantly impaired in the presence of TNFα (Fig. 3, A and B, and fig. S5). We next measured TNF receptor 1 (TNFR1) and TNFR2 expression levels because these two receptors bind to TNFα. Most (>90%) of the HSPCs expressed TNFR1, whereas only a small percentage (<10%) expressed TNFR2 (fig. S6). To assess the function of these two receptors, we used specific TNFR1- and TNFR2-blocking antibodies. Consistent with the receptor expression levels, blocking TNFR1 effectively abolished TNFα-induced reduction in CD34+ and HSC-e numbers, whereas blocking TNFR2 did not (Fig. 3, C and D). No additive effect was observed when both antibodies were introduced. In parallel, we tested the clinically available TNFα-neutralizing agent, etanercept (Enbrel, Amgen) (28), and as expected, its addition blocked TNFα-mediated effects. Collectively, these results, together with those from enriched HSC cultures (Fig. 2C), implicate TNFα as a potent inhibitory factor for the stem and progenitor cell compartment, consistent with previous studies (2932).

Fig. 3 TNFα inhibits CD34+ and HSC-e cell expansion via TNFR1.

(A and B) Fold expansion of (A) CD34+ and (B) CD34+CD45RACD90+ HSC-e cells from 7-day serum-free culture of UCB CD34+ cells in control or control + TNFα (1 ng/ml). Asterisk denotes statistical significance by two-tailed Mann-Whitney U test. Shown are means ± SD from nine independent experiments. (C and D) Expansion of (C) CD34+ and (D) HSC-e cells after 7 days. Cells were cultured in the presence of TNFα with the addition of etanercept, human immunoglobulin G (hIgG) control, TNF receptor (TNFR)–blocking antibodies (Ab), or mouse IgG1 control (5 μg/ml). Asterisk denotes statistical significance by Kruskal-Wallis test and Mann-Whitney U test with Bonferroni correction (Padj = 0.993) for comparing each test condition to the control. Shown are means ± SD from at least two independent experiments. *P < 0.05; **P < 0.01; and ***P < 0.001.

TNFα affects HSC survival and cell cycle progression and activates the nuclear factor κB pathway

For a deeper mechanistic understanding of TNFα signaling effects on HSPCs, we investigated apoptosis in our CD34+ cell cultures. In the presence of TNFα, the annexin-V+7-AAD (early apoptotic) and annexin-V+7-AAD+ (dead) cell frequencies increased in HSC-e within 24 hours, whereas the effect on more differentiated CD34+ cells was only detectable after 2 days (fig. S7). This suggested a direct mode of action of TNFα on HSC-e cells. Next, using time-lapse imaging, we continuously observed the most enriched HSC population (LinCD34+CD38CD45RACD90+CD49f+) (33) at the single-cell level for 5 days (movies S1 and S2). With TNFα treatment, 40% (28 of 70) of cells died. In contrast, 10% (8 of 77) of cells died in the control condition (table S6; P < 0.001 by χ2 test). The remaining living cells took significantly longer to undergo divisions in the presence of TNFα (median, 97.5 hours for generation 0 and 39.4 hours for generation 1) as compared to control (median, 54.2 hours for generation 0 and 23.9 hours for generation 1) (Fig. 4A). This was corroborated with a greater proportion of HSCs in G0 (Ki67 2n DNA content) after 2 days of culture with TNFα (Fig. 4B and fig. S8). Moreover, to interrogate the downstream responses of HSCs upon TNFα stimulation, seven intracellular signaling targets were measured using single-cell mass cytometry (table S7) (34). In LinCD34+CD38CD45RACD90+CD49f+ HSCs, TNFα uniquely activated the nuclear factor κB (NFκB) pathway as indicated by increased levels of phosphorylated p65 and decreased levels of IκB (inhibitor of NFκB), whereas the control condition activated different pathways including extracellular signal–regulated kinase 1/2 (ERK1/2), AKT, S6, signal transducer and activator of transcription 3 (STAT3), and STAT5 (Fig. 3, C and D, and fig. S9). However, when a pharmacological inhibitor of the NFκB pathway—Bay 11-7085 (10 μM)—was present, HSCs appeared stationary, and no cell divisions were observed (movie S3). This suggests a positive role for the NFκB pathway in counteracting TNFα signals. Together, these data demonstrate that TNFα leads directly to cell death and prolongs cell division time in human HSCs.

Fig. 4 TNFα directly induces cell death and prolongs cell cycle progression in HSCs.

(A) Continuous observation of single LinCD34+CD38CD45RACD90+CD49f+ HSCs by time-lapse imaging. Shown is time (hours) cells took to complete the first (generation 0), second (generation 1), and third (generation 2) division. A total of 15 cells did not divide by the end of the 5-day movie when cultured with TNFα. Data are from two independent experiments. (B) Ki67 and Hoechst staining were used to assess the proportion of HSCs in each phase of the cell cycle (G0, Ki67 2n DNA content; G1, Ki67+ 2n DNA content; S-G2-M, Ki67+ > 2n DNA content) over the course of 3-day culture of LinCD34+CD38CD45RACD90+CD49f+ HSCs. Shown are means ± SD of the percentage of cells in G0 from three independent experiments. Two-tailed Mann-Whitney U test was used to determine the significance level. *P < 0.05; **P < 0.01; and ***P < 0.001. (C) Heat map representation of the activation of seven intracellular signaling intermediates in LinCD34+CD38CD45RACD90+CD49f+ HSCs upon stimulation of control or TNFα only cultures by single-cell mass cytometry. Values are calculated as the difference of median arcsinh (signal intensity/5) of the indicated condition compared with the time 0 control and are shown in table S8. (D) Representative histograms of signal intensity of phosphorylated nuclear factor κB (pNFκB) and inhibitor of NFκB-α (IκBα) over time after stimulation of TNFα in LinCD34+CD38CD45RACD90+CD49f+ HSCs. Histogram color corresponds to the arcsinh difference calculated in (C).

Donor-derived memory T cells secrete TNFα

Given that donor-derived TNFα is elevated when HSPC engraftment is compromised, and that TNFα directly impairs the survival and division of phenotypically enriched HSCs in vitro, we next interrogated the direct effect of TNFα on donor HSPC engraftment and reconstitution in vivo. These experiments were carried out for 17 days at a maximum dose of 3 × 106 UCB cells, conditions where no significant weight loss and GVHD symptoms were observed (fig. S10).

To determine which donor-derived immune cell population(s) is(are) the source of TNFα, we used single-cell mass cytometry to measure the expression of five intracellular proteins (TNFα, IFNγ, IL-6, perforin, and granzyme B) and 23 surface antigens (table S9). Consistent with our initial observations, donor-derived CD4+ (36.8 ± 27.9%) and CD8+ (18.3 ± 4.5%) T cells were dominantly detected in the recipient BM when 3 × 106 UCB cells were transplanted. Comprehensive surface marker analysis allowed us to determine that most (>90%) of the CD4+ and CD8+ T cells expressed an effector memory (CCR7CD45RA) phenotype (Fig. 5A).

Fig. 5 Donor-derived activated memory T cells secrete TNFα.

(A) Representative spanning-tree progression analysis of density-normalized events (SPADE) analysis of human T cell subsets detected in the BM after transplantation of 3 × 106 UCB cells. The tree plot was constructed using 23 cell surface antigens. Shown are expression patterns of CD45, CD3, CD4, CD8, CCR7, and CD45RA. Putative cell populations were annotated manually by black lines encircling node sets expressing the indicated surface marker. The size of each node in the tree indicates the number of cells contained within each node. The color represents the arcsinh-transformed value of the global median intensity of the corresponding marker expression level of the cells in each node. CD3, CD4, and CD8 were used to cluster the T cell populations. CCR7 and CD45RA were used to further identify the naïve (CCR7+CD45RA+), effector memory (CCR7CD45RA), and central memory (CCR7+CD45RA) phenotypes within the CD4+ and CD8+ T cell populations. (B) SPADE representation of intracellular TNFα level in cell subsets identified as described in (A). Color scales represent arcsinh-transformed values of the global median intensity of TNFα expression levels, allowing direct comparison between samples (n = 2 mice per group from one experiment).

Significant levels of TNFα were detected when cells harvested from the recipient BM were subjected to a 4-hour pharmacological stimulation with phorbol 12-myristate 13-acetate and ionomycin (fig. S11B). The TNFα-producing population was clustered in a subset of memory CD4+ or CD8+ T cells that had an activated (CD38+HLA-DR+) phenotype (Fig. 5B). A significant IFNγ response was also obtained in the same cell subsets; however, we did not detect any IL-6 signal (fig. S11C). Note that the level of granzyme B expressed in the CD34+ cell population was 3 ± 1.2–fold higher in transplantations with unpurified UCB than with enriched CD34+ cells, suggesting that HSPCs experienced more stress (35) when mature immune cells were cotransplanted (fig. S11D). Together, this system-wide analysis of donor-derived cell subsets present in the recipient BM during the early posttransplantation phase reveals that the activated subsets of memory T cells are the major source of TNFα.

Etanercept enhances short-term HSPC engraftment and accelerates hematopoietic recovery

Capitalizing on the availability of clinically approved TNFα-neutralizing agents, we next investigated whether TNFα inhibition by etanercept would improve short-term HSPC engraftment numbers and hematopoietic recovery, a clinically important phase for HSCT (36). Infusions of the equivalent number of CD34+ cells were included as controls (Fig. 6A and figs. S12 and S13, A and B).

Fig. 6 TNFα-neutralizing agent enhances short-term engraftment of phenotypically and functionally defined primitive cells in vivo.

(A) Experimental scheme. i.p., intraperitoneally. (B) Percentage of CD34+ and CD34+CD45RACD90+ HSC-e cells detected in the BM after 17 days. Dots represent individual mice, with horizontal lines denoting the mean values (n = 6 or 8 mice per group from two independent experiments). (C) Heat map representation of the average percentage of human cell populations (gated on 7-AADCD45+HLA-ABC+) detected in the BM by day 17. (D) Heat map representation of human factor levels detected in the mouse serum after 17 days (from one experiment). Two-tailed Kruskal-Wallis and Mann-Whitney U tests with Bonferroni correction (Padj = 0.992) were used to determine the significance level. *P < 0.05; **P < 0.01; and ***P < 0.001.

Etanercept treatment significantly increased the percentage and absolute number of CD34+ and HSC-e cells detected in the BM of mice transplanted with UCB, compared to dose-matched no treatment or immunoglobulin G (IgG) control groups (Fig. 6B and figs. S13 and S14). Remarkably, numbers of CD34+ and HSC-e cells measured in the UCB + etanercept group were comparable to those obtained in the CD34+ group (fig. S13, C and D). Furthermore, etanercept affected the proportions of hematopoietic cell populations present in the BM, with a lower percentage of mature T cells (CD4+ and CD8+) and a higher percentage of progenitors (preNKT, proB, and MegaP) and mature cells of other lineages (Gran, Mono, and EryB), compared to the UCB control groups (Fig. 6C and fig. S15). Only in the etanercept treatment group were significant percentages of human myeloid (CD33+) and megakaryocytic (CD41+) cells detected by day 17 in the peripheral blood (PB) (fig. S16). The improved reconstitution number and diversity in the BM and PB are clinically crucial to achieve robust engraftment and faster hematopoietic recovery (36).

To confirm that donor T cells are the major source of TNFα in our model, we transplanted CD3+ T cell–depleted UCB and observed no significant difference between the etanercept and IgG control groups (fig. S17). Note that etanercept treatment led to changes in the composition of donor T cell subsets present in the BM, with higher percentages of the central memory and T helper 17 (TH17) subsets and lower percentages of the effector memory and TH1 subsets. A higher percentage of plasmacytoid dendritic cells were also detected with etanercept treatment (fig. S18). To verify that the increase in HSPC engraftment in the BM is not the result of systemic GVHD mitigation by TNFα blockade, we used a clinically effective GVHD-attenuating agent, tocilizumab (anti–IL-6 receptor antibody) (Fig. 7A) (22). Although we observed similar GVHD scores between the etanercept and tocilizumab treatment groups (fig. S19, F and K), blocking IL-6 (fig. S20) did not recapitulate the effect of blocking TNFα to increase the percentage and number of phenotypically (Fig. 7B and fig. S19, B and G) and functionally (Fig. 7C) defined progenitors engrafted in the BM and myeloid cells in the PB (fig. S19, C and H). Notably, the most stringently defined phenotypic HSCs (CD34+CD38CD45RACD90+CD49f+), which have been demonstrated to faithfully enrich for functional long-term HSCs after transplantation (37), were only detected in the BM with etanercept treatment (fig. S21). In addition, treatment of etanercept showed no negative effects on short-term (17 days; fig. S22) and long-term (12 weeks; fig. S23) reconstitution when CD34+ cells were transplanted. Collectively, these data further support the findings that blocking TNFα improves donor HSPC engraftment and reconstitution during the early phase of HSCT.

Fig. 7 Blocking IL-6 signaling does not recapitulate the effects of TNFα neutralization.

(A) Experimental scheme. (B) Percentage of CD34+ and CD34+CD45RACD90+ HSC-e cells and (C) total progenitor numbers quantified by colony-forming cell (CFC) assays detected in the BM after 17 days. PBS, phosphate-buffered saline. Dots represent individual mice, with horizontal lines denoting the mean values (n = 5 mice per group from two independent experiments). Two-tailed Kruskal-Wallis and Mann-Whitney U tests with Bonferroni correction (Padj = 0.992) were used to determine the significance level. *P < 0.05; **P < 0.01; and ***P < 0.001.

Next, to determine the effect of blocking TNFα on the production of other donor-derived inflammatory factors, we measured serum concentrations of human factors. Consistent with previous reports (38), etanercept treatment led to increased TNFα levels, suggesting that etanercept binding to TNFα prolongs its half-life in the serum. Other factor (IL-6 and IFNγ) levels were reduced with etanercept, suggesting that TNFα might act as a key initiating catalyst for the inflammatory factor secretion cascade (Fig. 6D and fig. S24).

Last, we observed that at increasing UCB doses, administering etanercept resulted in greater numbers of engrafted phenotypically enriched (fig. S25, B, E, and F) and functionally defined (fig. S25, C and D) stem and progenitor cells. Moreover, transplantation of 3 × 106 unpurified UCB with etanercept resulted in higher engraftment levels of HSPCs compared to transplantation of CD34+ cells isolated from 3 × 106 UCB cells. These data demonstrate the numerical advantage of using unselected cells in HSCT when combined with etanercept treatment by preventing cell loss associated with cell selection.

DISCUSSION

Recent evidence has suggested that HSCs can be activated directly by inflammatory factors under stress conditions (39, 40). Our study showed that inflammatory signals, generated by cotransplanted immune cells, can negatively affect donor HSPC function in a clinically relevant transplantation setting. We identified TNFα as a central player in cascading the “cytokine storm” that directly impairs donor HSC survival, division, and engraftment. We further demonstrated that neutralizing TNFα has two effects that are crucial in achieving successful post-HSCT outcomes (fig. S26). The first is an increase in the number and diversity of engrafted HSPCs in BM, which would ultimately promote stable long-term hematopoiesis. The second is an acceleration in the reconstitution of mature myeloid and megakaryocytic cells in PB, which facilitates timely hematopoietic recovery, thereby preventing neutropenia- and thrombocytopenia-associated complications (36).

The role of TNFα in regulating hematopoiesis has remained elusive. Several studies have shown stimulatory effects of TNFα in progenitor cell proliferation and reconstitution (41, 42), whereas other reports have implied that TNFα exerts suppressive effects in HSPCs (2932, 4346). There is evidence that the context-dependent effect of TNFα in culture is linked to the presence of other growth factors (47, 48). More recently, enhanced reconstitution of TNFR1 and TNFR2 double-knockout (Tnfrsf1-dKO) murine HSCs in serial transplantations has been demonstrated (49). Here, we provide evidence that the human UCB-derived HSC population is sensitive to TNFα signaling, directly prompting cell death and prolonging cell division time. This is consistent with recent observations that human HSCs have unique mechanisms to ultimately preserve HSC pool integrity in response to stress signals via selective apoptosis (50, 51) and regulation of cyclin-dependent kinase 6 (CDK6)–mediated G0 exit (37).

Aberrant production of TNFα has been linked to many malignancies such as acute and chronic inflammatory diseases and BM failure syndromes. In allogeneic HSCT, TNFα is elevated in the patient’s serum, and higher TNFα levels are associated with more severe GVHD and higher incidences of TRM (52). Our findings uncover an additional, previously underappreciated, role for TNFα in directly affecting donor HSC survival, division, and engraftment during the early phase of HSCT—a critical period for patient survival. The xenograft model we used in this study is limited in mounting recipient immune responses, and the development of GVHD after 17 days prevented us from assessing long-term hematopoiesis upon etanercept treatment. Nonetheless, clinical data from an early phase 2 trial suggest that the use of etanercept as part of a prophylaxis treatment correlates with lower TRM and higher 1-year patient survival rates (53). Further assessment of hematopoietic reconstitution in patients receiving etanercept as a preemptive therapy is necessary to evaluate the clinical relevance of the TNFα blockade approach. In addition, our system-wide mass cytometry analysis identifies the activated subsets of donor-derived memory T cells as the major source of TNFα, reminiscent of the phenotypic changes of T cells seen in the PB of patients within 28 days of UCB transplantation (54). Identification of antigens that stimulate donor T cell reactivity will aid in developing strategies to “design” a graft that enables the control of the signaling environment during the early phase of HSCT, by optimizing the composition of beneficial (normal or genetically engineered) T cell subsets (55) and specifically suppressing the undesired T cell phenotypes after transplantation.

In conclusion, our study establishes the role of TNFα in hematopoietic reconstitution in a clinically relevant HSCT model. Blocking TNFα signaling is a promising therapeutic strategy to protect donor HSCs and progenitors from the host’s detrimental inflammatory signals after transplant. The use of clinically available TNFα-neutralizing biologics may provide several advantages in transplantation of not only UCB but also other cell sources to (i) enhance stem and progenitor cell engraftment, (ii) accelerate peripheral hematopoietic recovery, and (iii) reduce substantial cell loss and costs associated with cell selection. These advantages will broaden the use of normal or genetically manipulated blood stem cells for the treatment of hematologic and nonhematologic diseases.

MATERIALS AND METHODS

Study design

Our objective was to characterize the effect of donor immune cell–derived inflammatory signals on transplanted HSPC fate in a clinically relevant transplantation setting. For in vivo transplant studies, no power analysis was performed to predetermine the sample size, animals were randomly allocated to treatment groups and randomly housed to minimize environmental factors that might confound the experimental outcomes, and endpoint analyses were performed in a blinded fashion. Sample size and number of replicates for each experiment are specified in figure legends. For in vitro cell culture experiments, cell samples pooled from at least two donors were used in individual experiments, and at least two independent experiments were performed.

Statistical analysis

Unpaired two-tailed Kruskal-Wallis and Mann-Whitney U tests were performed to test the statistical significance of the data due to non-normality assessed by the Shapiro-Wilk normality test. Bonferroni correction was applied for multiple comparisons as appropriate. We used R packages (version 3.0.2) for all statistical analyses with the α value set at 0.05 (unless otherwise stated).

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/9/421/eaag3214/DC1

Materials and Methods

Fig. S1. Significant weight loss was observed starting from day 17 when 5 × 106 UCB cells were transplanted.

Fig. S2. Human cell subsets detected in the mouse BM.

Fig. S3. Transplantation of UCB resulted in systemic inflammation as indicated by an array of human inflammatory factors detected in mouse sera within 2 weeks.

Fig. S4. Flow cytometry–based HSC-e cell assay.

Fig. S5. TNFα reduced the percentage of CD34+ and CD34+CD45RACD90+ HSC-e cells in culture.

Fig. S6. Flow cytometric analysis of TNFR1 and TNFR2 receptor expression.

Fig. S7. TNFα-induced cell death.

Fig. S8. TNFα inhibited HSC cell cycle entry.

Fig. S9. Intracellular signaling activation analysis by single-cell mass cytometry.

Fig. S10. Higher UCB dose led to reduced numbers of HSPCs engrafted in the BM, whereas no significant weight loss and GVHD symptoms were observed at 3 × 106 UCB cell dose.

Fig. S11. System-wide analysis of human cell subsets present in the BM by single-cell mass cytometry.

Fig. S12. Body weight and GVHD were monitored over the course of 17 days.

Fig. S13. Neutralizing TNFα increased the engrafted numbers of CD34+ and CD34+CD45RACD90+ HSC-e cells in the BM 17 days after transplantation.

Fig. S14. Numbers of human (CD45+HLA-ABC+) cell populations detected in the BM after 17 days.

Fig. S15. Human (CD45+HLA-ABC+) cell subsets detected in the BM 17 days after transplantation by flow cytometry.

Fig. S16. Etanercept treatment resulted in faster and enhanced reconstitution of human myeloid and megakaryocytic cells in the PB.

Fig. S17. Etanercept treatment had no significant effect on short-term engraftment of T cell–depleted UCB cells.

Fig. S18. Etanercept treatment led to changes in the composition of donor T and dendritic cell subsets.

Fig. S19. Blocking IL-6 receptor did not recapitulate the effects of TNFα blockade.

Fig. S20. Human (CD45+HLA-ABC+) cell subsets detected in the BM 17 days after transplantation by flow cytometry.

Fig. S21. Blocking TNFα increased the percentage of CD34+CD38CD45RACD90+CD49f+ HSCs present in the BM 17 days after transplantation.

Fig. S22. Etanercept treatment had no significant effect on short-term engraftment of human CD34+ cells.

Fig. S23. Etanercept treatment had no negative impact on long-term HSCs.

Fig. S24. Serum concentrations of 29 human factors with the treatment of etanercept.

Fig. S25. Dose-response experiment demonstrating the numerical advantage of using unselected UCB cells when combined with etanercept.

Fig. S26. Donor T cell–mediated TNFα signaling impairs the survival, division, and short-term engraftment of transplanted stem and progenitor cells.

Table S1. Cellular composition of the purified CD34+ fraction and unfractionated UCB cells.

Table S2. Numbers of CD34+ and CD34+CD45RACD90+ cells transplanted on day 0.

Table S3. Phenotypic definition of human cell subsets assessed by flow cytometry.

Table S4. P values determined by Kruskal-Wallis and Mann-Whitney U tests with Bonferroni correction (Padj = 0.992) for multiple comparisons among groups (corresponds to fig. S2).

Table S5. List of differentially overexpressed receptors in HSCs (the elevated serum factors are shaded).

Table S6. Data summary of time-lapse videos.

Table S7. Antibody panel for measuring intracellular signaling events using single-cell mass cytometry.

Table S8. Calculated median arcsinh (signal intensity/5) differences in stimulated CD34+CD38CD45RACD90+CD49f+ cells relative to unstimulated cells (at time 0).

Table S9. Antibody panel for measuring intracellular cytokine levels using single-cell mass cytometry.

Table S10. Antibody panel used to assess the human cell phenotypes in mouse BM by flow cytometry.

Table S11. P values determined by Kruskal-Wallis and Mann-Whitney U tests with Bonferroni correction (Padj = 0.992) for multiple comparisons among groups (corresponds to fig. S15).

Table S12. P values determined by Kruskal-Wallis and Mann-Whitney U tests with Bonferroni correction (Padj = 0.992) for multiple comparisons among groups (corresponds to fig. S16).

Table S13. P values determined by Kruskal-Wallis and Mann-Whitney U tests with Bonferroni correction (Padj = 0.992) for multiple comparisons among groups (corresponds to fig. S20).

Movie S1. Time-lapse imaging of human LinCD34+CD38CD45RACD90+CD49f+ HSCs in the presence of 3F.

Movie S2. Time-lapse imaging of human LinCD34+CD38CD45RACD90+CD49f+ HSCs in the presence of 3F + TNFα (1 ng/ml).

Movie S3. The inhibition of NFκB signaling immobilized HSCs.

Reference (56)

REFERENCES AND NOTES

Acknowledgments: We thank D. J. H. F. Knapp and C. J. Eaves for CyTOF technical advice, T. Chen for CyTOF protocol development and data acquisition, P.-A. Penttila and S. Zhao for fluorescence-activated cell sorting, and C. Yoon for technical support. Funding: Supported by the Human Frontier Science program (C.L.C. and P.W.Z.), the Leukemia and Lymphoma Society (P.W.Z.), the Canadian Stem Cell Network (P.W.Z.), the Canadian Institutes of Health Research (P.W.Z. and J.E.D.), the Ontario Institute for Cancer Research (J.E.D.), and the Terry Fox Foundation (J.E.D.). W.W. was supported by the Ontario Institute for Regenerative Medicine postdoctoral fellowship and the Swiss Initiative in Systems Biology Transition Postdoc fellowship. H.F. was supported by the Canadian Blood and Marrow Transplantation David Smyth fellowship, the Sears Cancer Research fellowship, and the Garron Family Cancer Centre Research fellowship (all the work by H.F. was performed at the University of Toronto). P.W.Z. is the Canada Research Chair in Stem Cell Bioengineering. J.E.D. is the Canada Research Chair in Stem Cell Biology. Author contributions: W.W. designed and performed all experiments, analyzed data, and wrote the paper. H.F., H.J.K., K.H., T.U., and Z.-J.L. designed and performed in vivo experiments, analyzed data, and edited the paper. S.X. performed the cell cycle experiments, analyzed data, and edited the paper. J.M. analyzed data and edited the paper. C.L.C., J.E.D., T.S., J.K., D.W., and R.M.E. provided funding, designed experiments, and edited the paper. P.W.Z. provided funding, designed experiments, analyzed data, and wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: CyTOF data are publicly available at Cytobank (https://sickkidsca.cytobank.org). Further inquiries should be directed to the corresponding author at peter.zandstra{at}utoronto.ca.
View Abstract

Navigate This Article